0:00

You and resident data science, but you're not

0:02

quite working in it yet. In.

0:04

Software Getting that very first job can

0:06

truly be the hardest one to land.

0:08

On. This Absurd. We have every smith from

0:10

Data career jump start here to share

0:12

his advice for getting your first date

0:14

a job. This. Is talk by

0:17

Than Me Episode Four and fifty Five Recorded

0:19

January eighteenth. Two. Thousand Twenty Four. Welcome.

0:36

To Talk By The Enemy a weekly

0:38

podcast on Python. This is your host

0:40

Michael Kennedy. Follow me on Mastodon where

0:42

I'm at Him Kennedy and follow the

0:45

podcast using at Tuck Python both on

0:47

Busted on.org. Keep. Up with the

0:49

show and listen. Over seven years have

0:51

passed episode at Top Python.fm. We.

0:53

Started streaming most of our episodes live

0:56

on you tube subscribe to our youtube

0:58

channel over at Talk Python.of them/you tube

1:00

to get notified of bout upcoming shows

1:02

and be part of that episode. This

1:05

episode is brought to you by century. Don't.

1:08

Let those areas go unnoticed. Use

1:10

Sentry like we do here at

1:12

TalkPython. Sign up at talkpython.fm slash

1:14

Sentry. And it's brought

1:17

to you by Posit Connect from the

1:19

makers of Shiny. Publish, share, and deploy

1:21

all of your data projects that you're

1:23

creating using Python. Pause.

1:31

It connects supports all of them.

1:34

Try Posit connect for free by

1:36

going to talk Python that Fm/posit

1:38

Pos Id. He before

1:40

we jump in and talk about data

1:42

science, jobs and careers. I want to

1:45

tell you really quickly about some awesome

1:47

news. Back in February I gave the

1:49

keynote app icon Philippines. It was entitled

1:51

the State of Python and Twenty Twenty

1:53

Four. Well. that is now

1:56

out on you tube that he met by

1:58

com philippines did a great job of came

2:00

out great. If you want to check out the state of

2:02

Python in 2024 according to me, just

2:05

click on the link in the show notes to watch it

2:07

over on YouTube. Now let's talk to Avery.

2:10

Avery, you're welcome to talk by the enemy. Thanks

2:12

so much. I'm so excited to be here and

2:14

be part of the show. I'm excited to have

2:16

you here as well. You know, one of the

2:18

things that people reach out to me often is

2:21

how you get into data science.

2:23

How do you get into programming? How do you

2:25

get into Python? You know, I've been trying

2:28

or maybe they got a degree

2:30

or they took some training program

2:32

boot camp or something and going

2:34

from zero to one I think

2:36

is the biggest career step you

2:38

have to make that next

2:40

job. And the one after that, it only

2:42

gets to be smaller steps, not bigger steps.

2:45

And it's really tough because that first big

2:47

step, you're brand new at it. You have

2:49

no experience, right? It's your first data science

2:51

job or your first programming job. So hopefully

2:53

we can give some folks out there a

2:55

little bit of a hand up to help

2:58

them make that jump. Yeah, totally. I like

3:00

to show this graphic that says

3:02

it's a circle and it's a circle of text.

3:04

And it says, I can't get

3:07

a job because I don't

3:09

have experience because and then it restarts,

3:11

I can't get a job. And that's the tricky

3:13

part. It's like, how do you get a data

3:15

science job when you have no data science experience?

3:17

Because to get data science experience, that seems like

3:19

you have to have a job as a prerequisite

3:21

and vice versa. So it is very tricky. So

3:23

I have to get to chime in on that

3:25

today. The industry can take it too far. They

3:27

can take it way too far. So a

3:30

few years ago, there

3:32

was a really funny tweet that went around

3:34

that back when they call them tweets. I

3:36

don't know what they're called anymore. Sebastian Ramirez,

3:38

the guy who created fast API, saw a

3:41

job posting when fast API was

3:43

like a year and a half old, it

3:45

said you must have four years of experience

3:47

with fast API to apply. He said, Hey,

3:50

look, I'm the creator of fast API. And

3:52

I'm unqualified for this job. What kind of

3:54

world are we living in? Yeah, that I

3:56

don't want to live in that world. where

4:00

we're at. That's so tough. And it's hilarious.

4:02

These job descriptions are getting out of hand,

4:04

that's for sure. Yeah, well, with AI, it's

4:06

probably not going to get better. We

4:09

could talk about that more later. But before we get into

4:12

that, you know, let's just jump in with a little bit

4:14

of background on you before we get to the topic. Tell

4:16

us a bit about yourself. What do

4:18

you do? How do you get into Python? Things

4:20

like that? Yeah, absolutely. So I'm currently a data

4:23

science consultant, and also a data

4:26

science instructor. I run some

4:28

online programs where I teach people to become

4:31

data analysts, mostly is what I'm focused

4:33

on. But I also have this practice

4:35

where I help companies solve data problems

4:38

with different techniques. I

4:40

started actually by studying chemical engineering

4:42

in college in my undergraduate degree.

4:44

And about a semester in I

4:47

realized crap, I hate this.

4:49

This is not for me. And I was a

4:51

little, a little bit taught you do you agree?

4:53

Have you felt something similar? I did a semester

4:56

of chemical engineering as well. I thought I love

4:58

chemistry. I love math, put them together. Somehow they

5:00

don't go together. It's like ice

5:03

cream and eggs or something. No, they don't go together

5:05

for me at least. Yeah, it wasn't good for me

5:07

there. I was just like, Oh, man, I'm actually like

5:09

not interested in refineries or like

5:12

manufacturing. But I like you like chemistry. I

5:14

liked math. I thought this is perfect. But

5:16

I quickly realized, oh, man, I really like

5:18

this whole programming part that I get to

5:20

do in MATLAB at the time, when

5:23

I was an undergrad, and I was on

5:25

a time crunch to get through college kind

5:27

of quickly through eight semesters. And

5:30

the other issue I had was I didn't know

5:32

what to do instead, it was like, I don't

5:35

really want to study computer science. Part of the

5:37

reason why is they can have this weed out

5:39

course at the beginning, which you had to build

5:41

Excel from scratch, basically, like some sort of a

5:44

spread sheeting tool. I was like, why would I

5:46

rebuild something that already exists that I don't even

5:48

like using in the first place, I wasn't really

5:50

into it. I didn't, I didn't know what to

5:53

do. And luckily, I was working as

5:55

a lab technician at this company, the

5:57

really cool company that makes the these

6:00

sensors that basically have the ability

6:02

to smell. So they can

6:04

sniff what's in the air and it has applications

6:06

for finding drugs or bombs and airports and stuff

6:08

like that. And there was a data scientist on

6:11

staff and that data scientist was awesome.

6:13

He was like showing me all these cool algorithms

6:15

he was writing for these sensors. And

6:17

then one day he got up and left

6:19

and he left the company and

6:21

we tried to hire another data scientist for

6:23

like six months, but they were really expensive.

6:25

We were a small company and none of

6:27

them really wanted to move to Utah and

6:31

so we couldn't really find someone that would be able

6:33

to do it. And so finally I was like, well

6:35

I really like this programming stuff and the data scientist

6:37

showed me a thing or two, maybe I could take

6:39

a stab at this. And I

6:42

wrote my first machine learning algorithm and I

6:44

was like, oh my gosh, I'm addicted to

6:46

this. And then I never looked back and

6:48

it has been data science sense basically. What

6:50

a great story. Yeah, I think a lot

6:52

of people fall into programming that way and

6:54

for some reason, not unexpectedly, but

6:56

for some reason a lot of people fall

6:58

into Python that way as well. They're like, you

7:00

know, I have a job and I got this thing I

7:02

gotta do. I just need a little

7:04

bit more than maybe like an Excel spreadsheet or

7:06

something and put it together and you're like, actually,

7:08

this is cool. After a while, like, this is cooler

7:11

than what I've been doing or maybe I'll make

7:13

it a good part of what I do, right? Yeah,

7:15

100%. Even just making, it

7:17

was in Matlab, which is basically engineer's

7:19

version of Python or college version of

7:21

Python 10 years ago, right? And I

7:23

made like tic-tac-toe and I remember playing

7:25

tic-tac-toe against the computer. I think that's

7:27

what it was or maybe it was

7:29

hangman. I can't remember. But I Remember like

7:32

the idea of like being able to play,

7:34

to program games and play against the computer

7:36

and I built it. I was like, this

7:38

is the coolest thing ever. I gotta do

7:40

more of this. Absolutely. You know, I think

7:42

I've done some Matlab too when I was

7:44

younger and it's not that different from Python,

7:46

but it's, I think one of the big

7:48

differences other than it just being like embedded

7:50

in a big expensive app is it's not

7:53

a general purpose programming language, right? You wouldn't

7:55

go, you know, that was fun, but let

7:57

me go build this website in Matlab or

7:59

let me... The Air B, M B

8:01

and Malabar you know like there's you, just

8:03

doors or is has this like self prescribed

8:05

a limit to what you can do with

8:07

it. That's one of the coolest part about

8:09

Python is it's really a Swiss army knife

8:11

and you can pretty much do and I

8:14

want to say anything but pretty close to

8:16

anything in Python which makes it really knee

8:18

and obviously one of the shooter limitations of

8:20

Mallayev is one it costs thousand dollars but

8:22

to your eye is not going to cyber

8:24

security for you. It's not going to build

8:26

web sites but the syntax at the end

8:28

of the day was was. Really quick it

8:31

was. It was easy for me a

8:33

transition from Malibu to Python cause the

8:35

syntax is another. The marriage is not

8:37

all that different, more math focus but

8:39

pretty similar to. I think maybe that's

8:41

a good place to start a discussion

8:43

exploring the topic of your first data

8:46

science job and witness. I plan us

8:48

are have a let's let's start with

8:50

before you even necessarily know programming language

8:52

right? Maybe you dabbled in Mount Lab,

8:54

or you dabbled in Excel or even

8:56

dabbled in on our javascript or somethin.

8:58

This thing we've been. talking about with

9:00

malibu and applies to other areas as well

9:03

like brew programming languages per se like a

9:05

julia or something like that is how if

9:07

you invest your time into learning one of

9:09

these things really well i guess how broadly

9:11

industrywide of of skill or high demand skills

9:14

are going to be right you are malibu

9:16

you put yourself in a box you are

9:18

in a more general programming language you can

9:20

have more options afterward drape yeah totally i

9:22

think like the more brought of a language

9:25

you learn the more useful you are to

9:27

to more industries in general that might take

9:29

that even a step further and just say

9:31

you know learning malibu not a whole lot

9:33

of companies use my lab but just like

9:35

landing your first date a job going from

9:38

zero to one is the hardest learning a

9:40

first language zero to one is artist as

9:42

well and then once you have that first

9:44

language the next language become so much easier

9:46

so one of first things i learned wasn't

9:49

was not lab and then i moved to

9:51

python and that was easier and i learned

9:53

sequel on that i learned our and that

9:55

i learned JavaScript and every time I added

9:57

like a new tool to my toolkit, it

10:00

was quite not I want to say easy,

10:02

but it got easier with each one. I

10:04

think that's true with foreign languages as well.

10:06

Once you learn one foreign language, then the

10:08

third and the fourth become quite easy. At

10:11

least that's that's what I heard. I speak

10:13

kind of two and a half languages, but

10:15

like there's people who speak like seven and

10:17

they always say like the six and the

10:19

sevens become easy. Yeah, you wonder how could

10:21

you possibly, because learning the first one is

10:24

so hard, first foreign language, so you're like,

10:26

well how could you possibly take that on

10:28

for this many languages and it's that

10:30

it's not the same challenges sign, right?

10:32

Yeah, exactly. Yeah, so I think when

10:34

people are considering getting into data science,

10:36

they really want to consider what

10:39

language they choose and where they go.

10:41

Like you're coming out of a college

10:43

program, you might feel like MATLAB or

10:45

something like that's real popular and yet

10:47

that's because it's popular amongst professors who

10:49

force their students to do it. That

10:51

doesn't necessarily mean that's the broad worldview.

10:54

What do you think about R? You

10:56

know, both. I like R. I'm not,

10:58

I sometimes troll R on LinkedIn. So

11:00

I guess that's another thing I should say is

11:03

I post a lot on LinkedIn, kind of a

11:05

LinkedIn guy. And so a lot of the times,

11:07

honestly, just for jokes and kicks and giggles, I'll

11:09

kind of roast R on LinkedIn just to get

11:12

the trolls angry in the comments and it's quite

11:14

fun. It's kind of fun experience, but I'm not

11:16

that big of a hater. I definitely think R

11:18

has its place. The one thing that's really interesting

11:21

about R versus Python and obviously a big debate

11:23

in the data science community is R is kind

11:25

of that does one thing really

11:27

well and it's getting a little less of that

11:30

as like more packages and libraries are added to R,

11:33

but R does the statistics and machine learning

11:35

very well. But obviously I don't think once

11:37

again, I don't know any websites, any like

11:39

super functioning websites that are built on R.

11:41

I don't know any cybersecurity that's really done

11:44

VR. So I think R does what it

11:46

does well. The syntax sometimes is a lot

11:48

easier for people to go from Excel which

11:50

a lot of people are more familiar with

11:52

in the finance or banking world. For example,

11:54

the syntax in R is a little bit

11:57

more similar to those Excel formulas than it

11:59

is to Python. So I

12:01

think sometimes people have a little bit more

12:03

success just because, oh, this kind of feels

12:05

like our formulas are sorry, this feels like

12:07

Excel formulas. And so people really get there.

12:10

I think what you're kind of alluding to

12:12

is if you're going to learn one skill,

12:14

you might as well learn the one skill

12:16

that's applicable to the most, the widest net,

12:18

right? And so that way you're fishing in

12:21

the biggest lake you possibly could versus in

12:23

a smaller pond of our I think that's

12:25

worth looking at. And one of the things

12:27

I actually really enjoy doing because you mentioned,

12:30

oh, you might think MATLAB is popular because that's

12:32

what the professors taught you. And there's actually not

12:34

a whole lot of data out there about, well,

12:36

what should you learn? So I don't know if

12:38

you know who Luke Baruch says. He's a data

12:40

analyst, YouTuber, I was going to say YouTuber on

12:43

YouTube, but that's kind of redundant on YouTube. And

12:45

one of the things he's done is he's

12:47

actually built this tool where he's web scraping

12:50

thousands of jobs, different data jobs every week,

12:52

and then displaying and analyzing the skills required

12:54

for those jobs. And it's actually like a

12:56

data driven way of saying, if you want

12:58

to be a data scientist, what

13:00

skills should you actually be focusing on as

13:02

you go, as opposed to just

13:05

listening to what a professor will

13:07

say, or what a LinkedIn influencer will

13:09

say or what your boot camp will

13:11

say like actually getting some data on

13:13

I think is pretty neat. That is

13:15

super cool. And I'm not familiar with

13:17

Luke. So we're going to dig him

13:19

up and put him in the show notes for later so

13:21

people can check that out. For sure. DataNerd.Tech,

13:25

I think is the website there.

13:27

I look at it mostly for

13:29

data analysts, because that's who I work with

13:31

the most. So I know the data analyst

13:33

data very well. SQL is number one at

13:35

50%. I think

13:38

Python is number two at like 30%. I

13:41

think Python might have jumped it. Well,

13:43

this is for all data positions right

13:45

here. So it's a job title you

13:47

can choose. Which one do you think

13:49

I should pick here? Data? Maybe

13:52

data scientists. Data scientists, yeah. Right.

13:55

Oh, wow. Python is 69%. Look

13:58

at that. That's Huge. Right? That's even.

14:00

That's even what Twenty Percent More Than Sequel was.

14:03

A lot of people are like is your amputees

14:05

scientists you ask know sepia if you were to

14:07

the job descriptions Pythons Benson does a lot more.

14:09

So if you're gonna learn as your brand new

14:12

and are going to learn one the muzzle start

14:14

with Python because it's probably the most in demand

14:16

skill that there is right now for the site.

14:18

yeah I'm is pretty easy rider saw like well

14:21

why don't you just learn C Plus Plus for

14:23

embedded devices are you know And maybe I'll pick

14:25

someone else to start with right? But a Python

14:27

pretty easy either. Easier? I think Pythons dry. it's.

14:30

I actually think I think sequels probably easier

14:32

to learn if I'm being honest as really

14:34

especially for like data science stuff is only

14:36

about twenty each. Man's a need a know

14:39

and sequel but it's once again sequels a

14:41

lot more. There's no website built on sequels

14:43

are you that much? so it's a lot

14:46

more limited on what agenda. it's his job

14:48

but not the language. It is not enough

14:50

on it's own. Generally Mn A you can

14:52

do reports and quite a bit with it,

14:55

but you know it's like when you see

14:57

these program popular at what's the most popular

14:59

language. A look see assesses the third

15:01

most popular. that's not a language. but

15:03

the thing, these with other languages right?

15:05

Like use it with all the other

15:08

languages that that's why it's high up.

15:10

But that doesn't mean as high in

15:12

demand exactly. It's just like table stakes,

15:14

you know? Yeah, telling our distinguish table

15:16

stakes from like picking an area. I

15:18

think that's totally true. And and really,

15:20

I think Python East as could make

15:22

the argument that there's really nothing in

15:24

sequel that you couldn't do in Python.

15:26

that's a little somewhat true depending on

15:28

data size and stuff like that. but

15:30

regardless there is ways that you can

15:32

do most of the sequel commands in

15:34

python one way or another down now

15:36

to be when when i first became

15:38

a data scientist i didn't even know

15:40

sequel and i was doing sequel commands

15:43

i was doing the aggregations or the

15:45

were functions or the window functions using

15:47

python solidify janet it's as long as

15:49

a date is not like super big

15:51

than you're totally right like some kind

15:53

of generator or even slices or ya

15:55

things like that right list comprehension second

15:57

projects are all that kind of stuff

16:00

kind of like, gosh, I really wish, a little

16:02

bit of a sidebar, but I wish like list

16:04

comprehensions and all those things had just

16:06

a few more SQL features, right? Like

16:08

in a list comprehension, I say, give

16:10

me this thing, maybe give

16:12

me this property of this class modified,

16:15

like give me the user's name uppercased,

16:17

right? So that's like select, and then

16:19

for thing in collection, that's like from

16:22

table or whatever, right? And then you

16:24

have the where clause with the if

16:26

statement, boy, wouldn't it be cool to

16:28

have like a sort also in

16:31

there and other

16:33

things like that, you know? Oh, well, totally.

16:35

It's so close. The cool thing is, is

16:37

if you want that sort, it's what, one

16:39

extra line? Like it's not too

16:41

bad. So Python, I mean, I

16:43

don't wanna say this necessarily to

16:45

make all the data scientists and SQL lovers mad,

16:49

but really Python can do a lot of the things

16:51

that SQL, that's for sure. Yeah, that's for sure.

16:54

This portion of talk Python to me is brought to

16:56

you by Sentry, code breaks, it's

16:58

a fact of life. With Sentry, you can

17:00

fix it faster. As

17:02

I've told you all before, we use

17:05

Sentry on many of our apps and

17:07

API's here at talk Python. I recently

17:09

used Sentry to help me track down

17:11

one of the weirdest bugs I've run

17:13

into in a long time. Here's what

17:15

happened. When signing up for our mailing

17:17

list, it would crash under a non

17:19

common execution pass, like situations where someone

17:21

was already subscribed or entered an invalid

17:23

email address or something like this. The

17:26

bizarre part was that our logging

17:29

of that unusual condition itself was

17:31

crashing. How is it possible

17:33

for her log to crash? It's

17:36

basically a glorified print statement. Well,

17:38

Sentry to the rescue. I'm looking at

17:40

the crash report right now and I see

17:42

way more information than you would expect to

17:44

find in any log statement. And because it's

17:46

production, debuggers are out of the question.

17:49

I see the trace back of

17:51

course, but also the browser version,

17:53

client OS, server OS, server OS

17:55

version, whether it's production or Q&A,

17:58

the email and name of the person. signing

18:00

up. That's the person who actually experienced the

18:02

crash. Dictionaries of data on the call stack

18:04

and so much more. What was the problem?

18:07

I initialized the logger with

18:09

the string info for the

18:11

level rather than the enumeration.info

18:14

which was an integer-based enum. So the

18:16

logging statement would crash saying that I

18:18

could not use less than or equal

18:20

to between strings and ints. Crazy

18:23

town. But it was sentry.

18:26

I captured it, fixed it, and I even

18:28

helped the user who experienced that crash. Don't

18:31

fly blind. Fix code faster with

18:33

sentry. Create your sentry account now

18:35

at talkpython.fm slash sentry. And

18:37

if you sign up with the

18:39

code talkpython, all capital no

18:42

spaces, it's good for two

18:44

free months of sentry's business plan which will

18:46

give you up to 20 times as many

18:48

monthly events as well as other features. Probably

18:52

the biggest, it's a bit of

18:54

a diversion, but the biggest similarity to

18:56

that I've seen in the languages is

18:58

C-sharp's link where they actually have almost

19:02

all the query operators including joins and stuff

19:04

like that built into the programming language. I'd

19:06

love to see more of that kind of

19:08

inspiration into Python, but you know, that's alright.

19:11

It's still really good. I've got a lot

19:13

of cool SQL-like features, but you're right. Once

19:15

you are no longer working with

19:17

data memory or you want indexes, right, like

19:19

this concept of indexes is not

19:22

sufficiently well understood, I think. Every time I hit

19:24

a website that takes five seconds to load, I'm

19:27

like, somebody is not doing all the things they

19:29

should be doing. I just know it. That's

19:32

totally true. What about SQL? You know, let's

19:34

talk about that for a bit, right? The

19:36

SQL, the query language, or

19:38

databases and other things. There's

19:41

ways to SQL query, not

19:43

just relational databases. But

19:46

you said you got away with not quite learning that, but

19:48

do you think if you could start

19:50

over maybe making an effort to learn that would

19:52

be really valuable? Like how important is this

19:54

a thing in the beginning of your career?

19:57

The interesting thing you know about landing a

19:59

data job... is your skills

20:01

only plays, I say, a third of

20:03

the role. Your portfolio or the way

20:05

that you portray your skills and

20:07

your network, I think, are the other two

20:10

thirds and they're actually more important than your

20:12

skills. And that's kind of how I got

20:14

away with not knowing SQL and not even

20:16

being, to be honest, that good at Python

20:18

at the time was because I used my

20:20

network to be in the situation to get

20:22

my lab technician job in the first place.

20:24

And then once again, I use that same

20:26

network, in this case, my coworkers, to land

20:29

that first data scientist position after we

20:31

couldn't hire anyone. And

20:33

if I would have been applying externally for

20:35

that role, chances are I wouldn't

20:37

have gotten that role. I probably didn't know enough

20:39

at the time to land that type of a

20:41

role. But because they knew I was hardworking, they

20:43

knew I wasn't like a total idiot and I

20:46

really liked to learn, they took that chance on

20:48

me. It paid off really well for them because

20:50

at the time I was still in college and

20:52

so I wasn't getting paid that much and

20:55

I was not getting paid like a data scientist,

20:57

and I was getting results like a data scientist

20:59

for them. So I think it paid off

21:01

for both of us. But I think if that was

21:03

an external job and I applied for it, I probably

21:05

didn't have enough skills for it. So I definitely think

21:07

learning SQL, if you wanna land a

21:09

data science job, isn't a bad place to start,

21:11

especially because like I said, I

21:14

mean, any programming language, I like to think

21:16

of like the icebergs, kind of like the

21:18

Titanic, right? There's the parts that you see

21:20

and then there's the parts that you don't

21:22

even know that are there. And

21:24

really you could spend the rest of your life

21:27

trying to master SQL or the rest of your

21:29

life trying to learn Python. But the cool thing is,

21:31

is a lot of the time you only need that top

21:33

little bit that's sitting at the top of

21:35

the surface of the water to actually get stuff

21:37

done. And so for SQL, I think that's like

21:39

20 commands and I think you

21:41

could learn it honestly in like a month.

21:44

You could learn those 20 commands pretty easily. But

21:46

it worked out for me and I didn't have

21:48

to use it that much at the time until

21:50

I was probably about almost three years into my

21:52

job and I actually had switched jobs to a

21:55

bigger company. The other thing that was working

21:57

for a smaller company where we didn't have a ton of

21:59

data. So we... use CSVs kind of as

22:01

our database, which is not great practice. But

22:03

when I eventually became a data scientist at

22:06

ExxonMobil, I was going to say they didn't

22:08

use Excel as a database, but they still

22:10

did. But the point is they had much

22:12

larger SQL databases with hundreds of thousands, actually

22:14

millions of rows of data that had to

22:17

query. Yeah, then you got to be really,

22:19

you need to understand it

22:21

at a much deeper level. Like if you do a query

22:23

like this, it's going to be super slow. But if you

22:25

do it like that, it can use the composite index for

22:28

the sort and then blah, blah, blah, blah, right, then you're

22:30

getting to the bottom of the iceberg and SQL

22:32

or maybe not the bottom, maybe like the middle

22:34

chunk under the water, but there's so much to

22:36

learn for both of them. Amira, the audience asks,

22:38

you know, like when you talk data job, like

22:40

what kind of jobs are out there, right? So

22:42

we talked to both about how we did chemical

22:44

engineering, and then we saw like chemical factors like,

22:46

yeah, I don't really want to work here anymore.

22:48

I'm out. So

22:51

thinking about like, well, what are the

22:53

kind of jobs you do? I think

22:55

that's really important because it's easy to

22:57

get focused in on like the same

22:59

companies. Like I want to work for like

23:01

some super big tech company, I want to

23:03

move to San Francisco and like that, that

23:05

right, like, there's not just plenty of other

23:08

jobs. But the opportunities just like you described

23:10

and as well as my first job, I

23:12

worked at a company that had like eight

23:14

people. And it was awesome, right?

23:16

They didn't expect me to be, you know, running

23:18

Kubernetes clusters and doing all sorts of great, they're

23:21

just like, I need you to make this thing

23:23

happen. Can you do like, I'm pretty new, but

23:25

that thing I can make that happen. Like, let's

23:27

go, right? And I feel like the the possibilities

23:30

to get in, especially with these maybe

23:32

more niche type of industries and companies might

23:35

even be easier for a first job. People

23:37

seem to be really obsessed with with the

23:39

saying, and I don't know if that's like

23:41

a societal thing, or if it's just those

23:44

are the companies that we use a lot.

23:46

And so we're excited about them. But yeah,

23:48

there's so many more data jobs outside of

23:50

fang than there are inside of fang, even

23:52

though there's there's quite a bit inside of

23:55

thing. And oftentimes, those roles can be much

23:57

more interesting, and you can do a lot

23:59

bigger. figure of an impact. When I

24:01

was working at this small company, Vaporsense,

24:03

I had so much power. I didn't

24:05

even realize it. I had such a

24:07

big effect on the company. I was

24:09

presenting to Fortune 500 companies,

24:11

and what I did really made a

24:14

difference. When it came to the point

24:16

where ExxonMobil offered me to go be

24:18

a data scientist for Exxon, I said,

24:20

oh, I want to go work for

24:22

the big company with the nice desk

24:24

and the nice laptop and try

24:26

something new. When I got

24:28

there, I had some pretty cool

24:30

opportunities when I was at ExxonMobil, but ultimately I

24:33

left pretty shortly after two years of being there

24:35

because I just felt like a cog in the

24:37

machine. I didn't feel like I was actually making

24:39

a difference. That was really important to my work

24:42

satisfaction of like, is what I'm doing being used?

24:44

Is it being used to better the world? Do

24:46

I feel valued? The answer was kind of no

24:48

for me when I was there. There's definitely a

24:51

tradeoff between the small companies and the big

24:53

companies, but also to go back to your original

24:55

question, there's so many freaking roles in the data

24:57

world that you're not even thinking of that I'm

24:59

not even thinking of. I saw a new one the

25:01

other day when I was helping one of my students. It

25:04

wasn't data janitor, but it was something like that where

25:06

I was like, I don't even know what that role

25:08

is, but there's so many roles.

25:11

When I was a data scientist,

25:13

Vaporsense, a small company, my actual

25:15

title was Junior Chemometrician, which basically

25:17

means you're doing data science with

25:19

chemistry. When I was at ExxonMobil,

25:21

when I was first there, I

25:23

was doing data science, but my

25:25

actual title was Optimization Engineer.

25:28

There's so many titles that we don't even think to

25:30

search of or even to look up, but

25:32

those are all data science roles. I was

25:34

doing machine learning every day in both those

25:37

roles, and you would maybe never guess from

25:39

those titles. Yeah, you would never guess. No,

25:41

that's awesome. What machine learning libraries frameworks were

25:43

you using? At Vaporsense, once again, because it's

25:45

a smaller company, I had a lot more

25:47

say in what I was doing. We were

25:49

building a bunch of machine... We were building

25:52

classification models to basically take the data from

25:54

our sensors and sniff if something was in

25:56

the air. Sometimes that was a yes, no.

25:58

Like, oh yes, there is a moan. in

26:00

the semiconductor factory and that's bad.

26:02

So that's a yes, classification kind

26:04

of binary, right? Other

26:06

times it was what drug is this?

26:09

Is this meth or is this heroin?

26:11

One of the use cases we had

26:13

was this is binary once again, but

26:15

is this recreational marijuana or medicinal marijuana?

26:17

And can we tell the difference between

26:19

those? So we are

26:21

usually using classification models usually

26:24

built in scikit-learn in Python the majority of

26:26

the time there. When I was at Exxon,

26:28

we had a lot less say, like

26:30

the data scientists had a lot less say

26:32

in the decision making process, we were doing

26:35

a lot of multivariate linear regression with a

26:37

lot of crazy hacks and transformations kind of

26:39

in the meantime for one of my positions

26:41

there. And then the other time, the other

26:44

position I did there, we were doing a

26:46

lot of auto ML using PyCarrot and letting

26:48

it kind of decide what type of models

26:51

to do. So. Okay, the

26:53

unsupervised learning type stuff, huh? It was awesome. It

26:55

was really fun to, I love

26:57

PyCarrot cause it's like, okay, go make

26:59

25 models and tell me which one's the best. It's like,

27:02

makes my job easy, I guess. We're

27:05

gonna be creative with sheer numbers. That's how we're gonna

27:07

come up with a solution. Got it. Exactly.

27:10

Diego is asking like, what are some of

27:12

the common stats methods as

27:14

in mathematical type stuff you would use? So

27:17

one of the things I know that

27:19

some people getting into programming think is you

27:21

gotta be really good at math to be

27:23

a programmer. I think you gotta

27:26

be really good at logical thinking, but

27:28

you need to almost zero math, be

27:30

like a web developer. You know, we're

27:32

talking percents for CSS, incrementing

27:34

numbers from one to two to two to

27:37

three for IDs and stuff like that. But

27:39

for data science, maybe there's a little bit

27:41

more. Like where do you see that kind

27:44

of background? I like it. You said you

27:46

have to think logically, but maybe the math isn't

27:48

as important. And I think it's actually somewhat similar

27:51

in data science. I will say you probably

27:53

need a little bit more math Than a

27:55

web developer, But I think it's a lot less

27:57

than most people think. And It's probably less about

27:59

being. What to do? The Math and

28:01

maybe more about understanding the mathematical concepts

28:03

and what I mean by that is

28:05

a lot of a lot of so.

28:07

I also have a master's degree in data

28:10

Analytics. A lot of masters degrees and

28:12

data Science Data Analytics will say you

28:14

need calculus and linear algebra has can have

28:16

a background for your mouth and not

28:18

that kind of stuff to plot on.

28:20

Julie Catalysts: I don't want union linear algebra

28:22

and while both those concepts do exist

28:24

and data science principles the majority of

28:26

the time, the computer I saw and

28:28

is doing the math use. As a

28:30

be able to interpret them, the results

28:32

of the math and encana know what

28:34

what different directions like this is going

28:36

down. One optimization problem. Yeah, okay, that's

28:38

the derivative you know the closer to

28:40

zero. That it's really less. About

28:42

knowing how to do the math by

28:44

hand and more does understanding what the

28:47

math the computer is actually doing. So

28:49

I think it's actually a lot easier

28:51

than most people say. That being said,

28:53

knowing how to do a derivative are

28:55

taking and are all those concepts I

28:57

think probably underlying pretty important. Burn that.

28:59

Like a lot of the times I'm

29:01

doing linear regression because it's It's awesome.

29:03

It gets the job done a lot

29:05

of the time I'm doing hypothesis testing

29:07

and statistics which also look like out

29:09

a Peace Corps nothing. all that crazy

29:11

attacks on. I did do a lot of

29:13

linear programming. But that's honestly that's like

29:16

the exception. Versus the rule. There's not a

29:18

whole lot of linear programming for most data

29:20

science most at a scientist so I really

29:22

don't think the masses is all that hard

29:24

Now of course as coming from someone who.a

29:26

chemical engineering degree who had to take all

29:28

the calculus all linear algebra so I did

29:30

go to Discourses. I had a really done

29:32

it from scratch from like a lot of

29:35

my students or teachers for example who never

29:37

took those courses in college So I can't

29:39

speak from that perspective but a lot of

29:41

my students are able to figure it out

29:43

into the day and insurance or so it

29:45

happens via for. Sir I think they're either

29:47

you by the it's about knowing okay

29:49

this formula or this algorithm or this

29:51

test means the seeing it applies in

29:53

a situation that doesn't apply in that

29:55

situation. Here's what you're trying to get

29:57

from it right? Like I know I

29:59

need to do a fast fourier transform

30:01

so and this is what it tells

30:03

me when I get out the other

30:05

side. But do I need to be

30:08

able to sit down and recreate the

30:10

integral and the that calculus behind it

30:12

and do that are like our how

30:14

are you the hallmark example I give

30:16

me a function of the the for

30:18

a transfer and I'll actually do the

30:20

symbolic integration like know you probably don't

30:22

need that right but you need to

30:24

know I do the for a transform

30:26

in this situation and this is why

30:28

and then I to say. Call.

30:30

The Function: Do it right and interpret

30:32

the results. Really? That's that's what being

30:34

a data scientist is all about His:

30:36

yeah, what does the business? You stays

30:39

with the desired business use case. How

30:41

I relate that use case to the

30:43

data what's technique. And I'd used to

30:45

guess the outcome that I need. Computer

30:47

Go Do it. Interpret results presents to

30:49

stakeholders. that's a data scientist Read: I

30:52

think one of the challenges with that

30:54

is gonna be not That is not

30:56

good. I think it's gonna be a

30:58

challenging because how do you learn. When

31:00

to use a certain statistical tests or

31:03

when the do some kind of funky

31:05

transformation like a for a a transform

31:07

without more traditional mathematical backgrounds and all

31:09

the academics will not just go. Always

31:11

going to give you like by men

31:13

are overview and I help you understand

31:15

their like know we're going to start

31:17

with this axioms or this theorem from

31:19

differential equation. that going to work up

31:21

like no no no no no I

31:23

don't need that, I don't I'm not

31:25

in a four year plan, I'm on

31:27

a four week plan. How to wait

31:29

and I. Get value from a couple of

31:31

the mathematical things without being sucked into like yeah

31:34

now I'm and differential equations at Harvard online and

31:36

I don't understand. I got there. It's such a

31:38

big problem and I'm so glad you brought this

31:40

up and I'll I'll be vulnerable the has yeah

31:43

I felt the same the same way and I

31:45

was like to ask me better way and so

31:47

that my was at three years ago. Now to

31:49

not years ago. three as you I said oh

31:51

my gosh I'm gonna solve this problem and I'm

31:54

gonna start my own data science to tip. And

31:56

so I spent about six months making the curriculum,

31:58

making all the videos. My. The ended up.

32:00

I got some students in there and I

32:02

ran for about six months. I look to

32:04

the results and man we weren't getting any

32:07

one into data science jobs and I thought

32:09

what the heck am I doing wrong I

32:11

had this really idea of like we're going

32:13

to be less theory, more project, more hands

32:15

on and I realized men the truth is

32:17

people to learn better at work. That's where

32:19

you learn that whole technique that would you

32:21

just like high and someone learned that the

32:23

answer is by getting experience and learning good

32:25

at work and when I looked back and

32:27

I said okay but we have had students

32:29

good. Jobs which opposite I guess and

32:31

it turns out most of them for

32:34

getting light business intelligence and intelligence Engineer

32:36

jobs or data analysts are financial analysts.

32:38

Jobs sour A little bit below a

32:40

data scientist shop, I realized oh man,

32:42

if we can just help people go

32:45

from zero to one and get their

32:47

foot in the door saying go from

32:49

one to five Much quicker at work

32:51

because work is just another is magical.

32:53

Place where it like like you said they did

32:55

you wherever you are working out earlier. Military can

32:58

you do this? To Brunetti saying this is to

33:00

throw you under fire. You'll figure it out and

33:02

that's somehow you do. I don't know what it

33:04

is about work for you. Figure it out and

33:06

that's where you learn. So that's kind of what

33:08

I've wide search my crew to and to be

33:11

more focus on it. Okay, maybe people are going

33:13

to become data scientist but we get them to

33:15

zero to one quickly and then they can get

33:17

paid to learn the rest of the day, decide

33:19

stuff when they're actually not first as as how

33:21

much you know about what you actually want to

33:24

do in the industry before you've done it as

33:26

well right? Like you're like, oh I thought everybody

33:28

said machine. Learning as awesome my view sets

33:30

u P C P D and I loved

33:32

it but it turns out actually like a

33:34

P eyes better. But I've never had a

33:36

chance to build an Ip I so until

33:38

I started I didn't even learn that one

33:40

was a thing to that. it was cool

33:42

or vice versa right? whatever. But into he

33:44

gets kind of in you don't even know.

33:46

like actually this part is where I really

33:48

am enjoying it and so just good enough

33:50

for stuff. That's a big deal hundred percent

33:52

you don't know what you don't know and

33:54

so you know it. That's why I mean

33:57

really when it comes to if we'd. Like

33:59

to go back to just see. Or just I thought

34:01

you could spend. I tell people this, if

34:03

you tried to master Python before you applied

34:05

to a job, you'd be like eighty years

34:07

old before you ever applied to a jot

34:09

down. Same a sequel. Same with machine learning.

34:11

The cool thing about data is we're never

34:13

going to know at all and so just

34:15

learned the bare minimum to get your foot.

34:17

In the door and then you have this place

34:19

where user and get paid to learn what you

34:21

want to learn eventually as he learned oh I

34:23

love a T I's I promise you that there's

34:25

a company out there that will hire. You and

34:27

you can learn a p eyes on the job

34:30

like that's joint app and but that first episodes

34:32

who's a company out there that it doesn't know

34:34

it needs a pure Css but you can help

34:36

them and your They don't have a huge expectations

34:38

because this is a thing they just learned they

34:40

needed right onto person from has its while right?

34:44

This. Post natal python is brought you

34:46

buy posit the makers of Seine

34:48

formerly our studio. And especially

34:51

signing for Python. Let.

34:53

Me: ask you questions. Are you building

34:55

awesome things? Of course you are. You're

34:57

a developer, data scientists. That's what we

34:59

do and you should check out Posit

35:01

Connects. As a Connect is a

35:03

for you to publish sheer and deploy

35:05

all the data products that you're building

35:07

using Python. People. Ask me

35:09

the same question all the time. Michael I have

35:12

some cool data science project or nope of that

35:14

I built. How do I share

35:16

with my users stakeholders teammates? Or.

35:18

Need to learn fast? A Pr

35:20

flask or maybe viewers react is.

35:23

Over now those are cool technologies and I'm

35:25

sure you benefit from them, but maybe stay

35:27

focused on the data project? Let posit

35:29

connect handle that side of things. With.

35:31

Posit Connect You can rapidly in

35:34

securely deployed the things you build

35:36

in Python Streamer-shiny book if fast,

35:38

A P Flask, Quarto Ports, dashboards

35:40

and a P Eyes. Posit:

35:43

Connect support or them in posit.

35:45

Connect comes with all the bells

35:47

and whistles to satisfy I T

35:49

and other enterprise requirements. make

35:51

deployment the easiest step and york flow with

35:53

posit connects for limited time you can try

35:55

posits neck for free for three months by

35:58

going to top python been a family That's

36:01

talkbython.fm slash P-O-S-I-T. The link

36:03

is in your podcast player

36:05

show notes. Thank you

36:07

to the team at Posit for supporting TalkByThon. All

36:13

right. Let's talk about some career advice.

36:15

I mean, I know you talked about

36:17

being connected on LinkedIn pretty well and

36:19

certainly having some kind of social network

36:21

of support. And they maybe, it's not

36:23

that you would call it not social,

36:26

but a real world network of actual

36:28

human beings that you physically

36:30

know somehow. What's that? I don't know

36:32

what that is. I know. We

36:34

gave that up back in 2020, I thought. Yeah. Anyway,

36:36

there was some stats that I saw somewhere that over

36:39

half of the jobs are fulfilled before even

36:43

becomes a job posting. Maybe some

36:45

of the best ones is like, hey, who knows somebody

36:47

who can do this? Your

36:49

data scientist example, they quit. We

36:52

need somebody, anybody know a good data scientist. I

36:54

don't want to just go put it out on

36:56

the open job market and have to have a

36:58

hundred interviews and who knows what I'm going to

37:01

get. If you can recommend somebody, let's start

37:03

there. So being in that

37:05

group to be recommended is important. It's

37:07

the key. There was a really

37:09

interesting survey done on LinkedIn and they said, it

37:11

was done by the same person

37:14

and Jordan Nelson, by the way, he said,

37:16

how do you approach getting a job? And

37:18

then the next day he said, how did you get your

37:21

last job? And 80% of people,

37:23

they use what I call the spray and pray

37:25

method, which basically means you go and you apply

37:27

to as many jobs as you possibly can and

37:29

hope for the best, cross your fingers. That

37:31

was 80% of what people were doing. And

37:33

then on the next poll, the next day, it was a

37:35

total, I think of what, 70% were either head

37:39

hunted, recruited or referred. Yeah. And

37:41

so it's like the Pareto principle here where

37:43

80% of the effort is only getting you 20% of

37:45

the results. And

37:48

really 20% of the effort gets 80% of the results. So

37:50

it's okay, we know networking and getting recruited is

37:53

really important, but how do we do it? It's

37:55

easier said than done. And Like you said, in the

37:57

industry, how do I make friends who are right? Oh

38:01

my neighbors don't do it's I guess I'm out.

38:03

That's the tricky thing is is yeah through Nine

38:05

industry yeah How do you get recruited into our

38:07

how do you know someone and will have come

38:09

to learn. Is it actually doesn't even

38:11

matter? Say. For instance, let's take a

38:13

seat, your neighbor, right? Your neighbors probably

38:15

not a data scientist. maybe you're lucky

38:17

and they are and they can refer

38:19

you to accompany the what's really cool.

38:21

As I've learned that companies really come

38:24

to trust their employees and their employees

38:26

are recommendations. And so even if your

38:28

neighbor let's say is a web developer

38:30

or maybe even less technical a say

38:32

your recruiters and finance right? If there's

38:34

an opening, I get data science opening

38:37

at that company. a law that times

38:39

they will actually take their employee referrals

38:41

much more. Seriously than any sort of cold

38:43

application that they get. And so other times

38:45

I've had students who just know someone that

38:48

works at the company they saw a job

38:50

opening pop up there quickly. They messes their

38:52

friends hey, you know, recruiter or a hiring

38:54

manager. I could talk more about this role,

38:56

but you do an internal referral. For me,

38:59

and they're able to land jobs that they probably

39:01

wouldn't have. No, the definitely wouldn't have

39:03

without that internal referral so it is

39:05

tricky. It's the old cliche, it's not

39:08

so what you know, it's a dinner

39:10

still player was covered notwithstanding. I think

39:12

that these days is plenty way sir

39:15

get those connections right? But maybe people

39:17

are know like meet up.com is really

39:19

good if you live in the non

39:21

tiny city. there's many many things going

39:24

on that around data science around I

39:26

thought around other data engineering whatever or

39:28

you could go to. Those things are

39:31

typically. Even free. often they have

39:33

free with food. the when feed

39:35

your rights and make connections or

39:37

regional conferences or national conferences re

39:39

like we probably many people voted.

39:41

Bike on right? There's us Python,

39:43

There's Euro Python and then there's

39:45

but that's those are the ones

39:47

that are lost and talks about.

39:49

But there's. send money little smaller

39:51

regional ones in the us and many more

39:54

them not aware of throughout the world probably

39:56

one of those within driving distance right that

39:58

you could go to me connections and

40:00

just also kind of take the temperature

40:03

of actually what you see on the internet

40:05

versus what you see actually talking to real

40:07

people. So I'd also

40:09

say just get out there. A hundred

40:11

percent. Those places have the people who

40:14

probably want to hire you because

40:16

they're local, right? Which is one thing that's

40:18

trouble on LinkedIn. I'm big on networking on

40:20

LinkedIn. But a lot of the times, you're

40:22

going to be networking to people who in

40:24

all likelihood might never have a role that's

40:26

even open to you. But the people that

40:28

you're like, for instance, we have, I'm

40:30

in Utah, we have Silicon Slopes that

40:33

has like a tech meetup. We have

40:35

a local Python meetup chapter. We have

40:37

the big data and developers conference that's

40:39

free every year with tons of food.

40:42

And the people who go there are people

40:44

from companies around there that have the openings

40:46

that you're trying to find. And they want

40:48

to hire people like you who are in the

40:50

area. So at least you can maybe come to the office

40:53

once a week or maybe once a month or whatever,

40:55

right? Like you said, going

40:57

to those meetups, it's tough because networking

40:59

is always difficult, either online or in person.

41:02

But at least in those situations, you know,

41:04

hey, these are people that are tied to

41:06

real companies that exist around me that do

41:08

make data hires. So I have a chance.

41:10

Definitely a much higher chance than just shooting

41:12

out a resume. All right, let's see. We

41:15

talked about job hunting already. What

41:17

about like applications and resumes? What are

41:19

your thoughts on that? I think once

41:21

again with the applications, the more targeted that

41:23

you can make it, the better, right? So

41:25

if you can really hone in on I

41:28

really want this job, I'm going to cold

41:30

message five people at this company and see

41:32

if I can get that internal referral one

41:34

way or another, make a real connection with

41:36

them. I think that's really key. And

41:39

then with resumes, resumes are more of an

41:41

art than they are a science. I feel

41:43

like like they are so difficult to figure

41:45

out. And these ATS is that are

41:47

trying to match you and see if you're a good

41:49

fit. I've tried a lot of them and a lot

41:51

of them suck whoever is the data scientist behind

41:54

those. We need to have a conversation with them

41:56

because it's a little tricky sometimes. But one

41:58

of the coolest concepts I've been interested in is the introduced

42:00

to recently, and I have a

42:02

whole episode on my podcast about

42:04

it, is A.B. testing your resume.

42:06

And basically, the idea is a

42:09

resume's job is just to get you a

42:11

screener interview or like a beginner interview basically,

42:13

right? That's all an interview. Like no one's

42:15

seeing a resume and then hiring you. They're

42:17

always going to interview. So if you think

42:19

about it, a resume's job, the only job

42:21

it has is to convince someone to get

42:24

on the phone and talk. And

42:26

it's just a piece of paper. And guess what? You

42:28

can put whatever you want on that piece of

42:31

paper. Now I'm not saying to lie, but I'm

42:33

just saying you could theoretically make a perfect resume

42:35

for whatever job you're trying to go for and

42:37

send it out there and see what happens, right?

42:39

But I'm not saying to do that. I'm not

42:42

saying to lie. My point in saying this is

42:44

that the resume is just to get you the

42:46

interview. And if you're not getting interviews, something's probably

42:48

wrong with your resume. And so

42:50

tweak something, apply 10 more jobs,

42:52

see what happens. Tweak something,

42:54

apply 10 more jobs, see what happens. Until

42:57

you finally have the right combination, skills,

42:59

of experiences, of different keywords, because a lot

43:01

of the time you're just trying to beat

43:03

the ATS. And that's the sad part

43:06

about it is it's like, how do I prove to

43:08

this random computer algorithm that they should talk to me

43:10

on the phone? That's a hard game to beat. And

43:13

there's a whole bunch of advice from all these

43:15

different people. What I've come to learn is it's

43:17

different for every company. It's different for every person.

43:20

You kind of a numbers game till you get lucky and

43:22

you figure it out. That's good advice. I

43:24

guess two thoughts. One is I know

43:26

that speaking specifically to anyone,

43:28

but in general, women wait until

43:30

they match all the requirements of

43:32

a position where a guy's like,

43:35

I know three of those things. I'm taking

43:37

a flyer, I'm sending it. I would just like

43:40

to encourage the women out there to just send

43:42

it as well. I 100% agree with that. And

43:44

I think if you reach 60% of

43:47

the requirements, I think you have a chance.

43:49

Like it's a lot of the times those

43:51

are wish lists and not actual requirements. And

43:53

depending on are you local to the area?

43:56

Do you have a domain experience in

43:58

this company? There's lots of other. What

44:00

about contributing to open source or

44:03

having GitHub repos that can be like

44:05

projects that you can show off or

44:07

what's your advice there? I'm a huge

44:09

proponent of projects in a portfolio. I

44:11

think if you don't have experience with

44:13

something, you create your own by building

44:15

a project. And if you can

44:17

do that with open source, I think you should

44:19

totally do that because I've benefited so much from

44:21

open source. I have not given

44:23

back as much as I should to open

44:25

source development and projects. I definitely should do

44:28

that. But if you can find a project

44:30

that you're passionate about, that you can help

44:32

with, I think you should totally do that.

44:35

Even if it's not open source and you're just

44:37

building a project to showcase your skills, I'm all

44:39

about that. I think you can do projects that

44:41

are super fun, maybe that are good for your

44:43

community or good for your life. I'm a huge

44:45

fan of personal projects. I've put a

44:47

Fitbit on my dog before and looked at

44:50

her steps. I've found the healthiest meal at

44:52

McDonald's. I've looked at like visualizing my weight

44:54

over time and tried to create like different

44:56

like forecasting models and stuff like that. There's

44:58

so much data in our lives that you

45:01

can use to make really cool projects. No,

45:03

absolutely. You talked about, okay, you get your

45:06

first job and that's where you kind of

45:08

really learn. But if you don't have your

45:10

first job, you can effectively simulate that. Say

45:12

I would have gone to a job and

45:14

been given a project to analyze something. I'm

45:16

just interested in this thing. I've got two

45:18

hours a day until I get a job

45:20

that I can be inspired about this and

45:23

just get going on it. Maybe create a

45:25

website and publish your results and it can

45:27

draw more people in to actually see that,

45:29

right? And start to appreciate them. They could

45:31

even ask like, all right, who's behind this

45:33

cool project? Maybe I want them

45:35

to come work for me. Little did they know

45:37

you're doing all this work because you got some

45:39

spare time and you're trying to build up your

45:41

experience and a self-guided study, right? Yeah,

45:44

if you can build a cool project and flip the job

45:46

hunt where you're not applying for jobs, but

45:48

jobs start to apply for you, you're in

45:50

such a good position. And doing really cool

45:52

projects can help you get there. Now,

45:55

it's hard to do cool projects. It's

45:57

hard to publish projects, which is one of the things that

45:59

people really strive for. struggle with. For all

46:01

you Python listeners out there, let

46:03

me just tell you, Streamlit is

46:05

absolutely amazing because it makes the

46:07

deployment process so easy.

46:10

It's free. It's a

46:12

little tricky to deploy at first, but compared to

46:14

what you used to have to do back

46:17

in the day, I'm saying back in the

46:19

day, like four years ago, basically. But it

46:21

was really hard to deploy something where you

46:23

could send someone a URL, hey, check out

46:26

my web application, learning application. Streamlit

46:28

is such a cool app that makes it so easy

46:30

and so intuitive to

46:33

make these cool little apps that you could just put on

46:35

your resume, put on your portfolio, send

46:37

to recruiters. I'm such a fan of the Streamlit

46:39

app. I love it. Yeah, it's super cool. There's

46:41

a couple of those and Streamlit is definitely one

46:43

of the really nice ones there. They host it.

46:46

There's like some hosting behind Streamlit as well these

46:48

days, right? You don't even have to set up

46:50

a server or anything. You just create it and

46:52

put it up there. I think what I'm saying

46:54

is back in the day, I use

46:57

Dash a lot. I'm still a big fan

46:59

of Dash. Dash is more customizable than

47:01

Streamlit and can do quite a bit

47:03

more, but it's a lot

47:05

more work to deploy it. More like programming. Yeah,

47:07

it is more a lot more programming. It's more

47:09

like a UI rather than just the behind the

47:12

scenes. You have to do both.

47:14

You have to know a little bit about

47:16

systems and data engineering and stuff like that

47:18

versus Streamlit kind of takes that, abstracts that

47:20

away. Back in the day, I

47:23

used to make Dash web applications and deploy them

47:25

on Heroku back when they had a free tier

47:27

of hosting and they've taken that away. I don't

47:29

even know what the go-to free hosting

47:31

platform is nowadays. I moved most of my things

47:33

to Streamlit and it's so nice. We got Shiny

47:35

for Python now, which is also nice. I haven't

47:38

checked that out. How is it? I haven't done

47:40

too much with it either, but Joe

47:42

and the team over there are doing pretty cool

47:44

stuff like adding more dynamic interactive stuff to Jupiter,

47:46

like running it inside Jupiter and things. Yeah, pretty

47:49

cool. I'll have to check it out. I think

47:51

they also do a bunch of hosting stuff over

47:53

there as well. That's why it came to mind.

47:55

What other advice do you got for folks out

47:57

there? So AI is AI. not

48:00

studying AI or learning to use AI and

48:02

machine learning, but is there a benefit of

48:04

trying to, you know, like use chat's GPT

48:07

to help you get this job? Or is

48:09

there a danger? Like I'm thinking, for example,

48:11

like have a chat GPT write me an

48:13

awesome resume. And then the tools are like,

48:16

well, we've detected this is AI generated, and

48:18

it's out. You know what I mean? Like,

48:20

what do you see happening there? A lot

48:23

of people see AI as like an all

48:25

or nothing tool, as in it's, it's either

48:27

you the human doing the work, or it's

48:29

the AI doing the work. But whenever I

48:32

don't know about you, but whenever I'm using

48:34

chat GPT for anything, it's very rare. It's

48:36

copy and paste for me, or at least

48:38

not iterative, where I'm doing multiple prompts, prompt

48:40

after prompt after prompt trying to tweak it

48:43

exactly what I want. And so the way

48:45

I look at chat GPT and other gen

48:47

AI that will be coming out, that's only

48:49

inevitable is instead of looking at does this

48:51

replace me? Does this like, for instance, am

48:53

I going to build my whole resume using

48:56

chat GPT? Am I going to build it

48:58

can chat GPT build, you know, take a

49:00

data scientist job and build the whole

49:02

model for them? I like to see

49:05

it more as like a hammer. It's

49:07

like a tool for the data scientist

49:09

or a tool for the job searcher

49:11

to using conjunction with your screwdriver or

49:13

it's like something to be wielded by

49:15

human not replaced for the human. That

49:17

makes sense. You know, it's really good for stuff

49:20

like, hey, there's I know a regular expression will

49:22

do this. Yeah, the last time

49:24

I studied, I completely forgot what this is about.

49:26

And I know it's gnarly. But if I just

49:28

ask, here's an example, here's what I want. Boom.

49:31

And traditionally, what you would end up doing is you'd

49:33

be on Stack Overflow. Yeah, be all

49:35

over the internet, you'd be trying to piece it

49:37

together from external information anyway. And so code

49:40

is something that's a little bit more

49:42

in the wheelhouse of the generative AI, because

49:44

it can't really make it up as much.

49:46

I know it could like do something

49:49

insecure, and you didn't know it was or whatever.

49:51

But it's not like asking for legal

49:53

advice where it makes up cases that didn't exist.

49:55

Like it gives you code, you put it in

49:57

the runtime for the compiler and it

49:59

runs. or it doesn't, the output comes

50:01

like you did. Yeah, it works or not.

50:03

Yeah, so it's pretty effective for that. But

50:06

yeah, for resumes, I would be more like,

50:08

let me ask it, what are the in-demand

50:10

things? And if I know these three skills,

50:12

what other skills should I know to get

50:15

a, like you could sort of use it

50:17

in an explorative way to then come up

50:19

with what you might write for yourself, right,

50:21

something like this. I find it really useful

50:24

for brainstorming like action verbs on your resume

50:26

bullets. Like I think it's really good at

50:28

that. What's 10 different ways to say led? So

50:31

I don't say led five times on my resume

50:33

and I use some different action bullets. I think

50:35

it's great at that. I personally,

50:37

it's pretty rare that I start

50:39

any Python code from scratch nowadays.

50:42

I'm either starting hopefully from a template

50:44

that I've already written, or

50:46

I'm starting from a chat GPT. Like

50:48

this is what I kind of want to accomplish, write

50:51

like the outline for it. Like one of the things

50:53

I hate doing is I make a

50:55

lot of streamlet apps. I probably make a streamlet app a month right

50:57

now. And I hate starting from scratch with

50:59

streamlet. It's super easy to start from scratch, but I'll

51:01

say, hey chat GPT, I want to build a streamlet

51:03

app. This is like the component I want here. This

51:05

is the component I want here. This is the component I want here.

51:08

And it's almost like a warmup for me

51:10

as a programmer and it will create something

51:12

that works. It's not what I want. And

51:15

I spend the next five hours trying to

51:17

make it what I want, you know, without

51:19

chat GPT, but it kind of gives me

51:21

a warm start to my programming process. So

51:24

I really like it. I think it's something that

51:26

everyone should use. And I think if you're thinking

51:29

about getting into any sort

51:31

of programming, you know, whether it's data

51:33

science or web development, I think you

51:35

should be a little bit less worried

51:37

about it taking your job and job

51:40

security. I think you should almost

51:42

be more excited that, wow, the bar has

51:44

never been lower to break into tech. Like

51:46

this is a step up gift from the

51:48

programming gods that I get to use to

51:50

break into tech. Another thing to keep

51:53

in mind is I imagine a lot

51:55

of people listening to this podcast are

51:57

not just starting a college program, right?

51:59

They're coming from... possibly other experiences,

52:01

other specialties. You know what's really

52:04

good for job security? Knowing the

52:06

intersection of two things. The intersection

52:08

of chemistry and programming. The intersection

52:11

of geology and programming for Exxon

52:13

potentially, right? Like those things take

52:15

you from a pool of 1,000

52:18

to a pool of 10s, right? And so what's

52:21

awesome about that is it means two things. You

52:23

don't throw away, if you got a degree in

52:25

something else like biology or whatever, you don't throw

52:27

away like, well that was wasted four years, that's

52:29

out. And it slices the pool

52:31

of people who could apply for certain jobs

52:33

way, way smaller, right? Sounds like you agree.

52:35

Oh, 1,000%. I'll

52:38

just tell a quick little anecdote. When I was

52:40

at ExxonMobil, there's a lot of things I did

52:42

not like at ExxonMobil. But this is something I

52:44

really liked. It's about once a

52:47

quarter, they would do a crowd-sourced data science

52:49

competition for the whole organization, like around the

52:51

entire world. And they would say, this is

52:53

a business problem we're trying to solve. And

52:56

at Exxon, we have data scientists all over the

52:58

world in all sorts of different teams and things

53:01

like that. So I did not know all the

53:03

data scientists at Exxon. And they'd say, this is

53:05

the problem we're facing, here's the data, go, right?

53:08

Nice. And I loved participating in

53:10

these. It was right up my wheelhouse of, I

53:13

really enjoy exploration and all this stuff. At the

53:15

time, I was getting my master's degree, but I

53:17

didn't have my master's degree. And I was competing

53:19

against, so I'm just a

53:21

chemical engineering grad, right? And I'm competing

53:24

against people with PhDs in computer science

53:26

and in data science and all these

53:28

people who have way more experience than

53:30

me. And I actually won a few

53:32

of these competitions. Thank you, I appreciate

53:34

it. And it's not because I was

53:36

a better programmer or a better data

53:38

scientist, it's because I majored in chemical

53:41

engineering. And I knew the business problem,

53:43

the domain extremely well. And I kind

53:45

of knew the programming and the data

53:47

science stuff, but the combination of them

53:49

made me very valuable. Like one of

53:51

the best examples I have is, we

53:54

were looking at crude oil properties. And

53:56

I remember there was a forum where

53:58

you'd ask your questions. One of

54:00

the data scientists asked, hey, is sulfur bad?

54:02

There's lots of sulfur in this, is it

54:04

bad? And like to a chemical engineer, that's

54:07

like the most obvious thing. No, yes, sulfur

54:09

is very bad in crude oil. That's very

54:11

no-no. That's like such a fundamental thing to

54:13

me and to him or

54:15

her that was like groundbreaking. And so

54:18

yeah, your domain can become your superpower

54:20

in your career. And it makes it

54:22

way harder for chat, GPT and other types

54:24

of tools to just automate you out of

54:27

a job because you bring in all these

54:29

skills together, which is awesome.

54:32

It makes it easier for you to continue

54:34

your momentum of whatever you've been up to.

54:36

It's just, it's good all around. Yeah, I

54:38

think it's more fun too, because once again,

54:40

like when I was trying to decide if

54:42

I should study computer science, I was like,

54:45

man, I don't really want to build an

54:47

Excel workbook for building an Excel workbook sake.

54:50

That's still true for me today. I don't

54:52

wanna do data science for data science sake.

54:54

I only like machine learning or data science.

54:56

When I'm doing it to solve a really

54:59

fun problem, I'm passionate about

55:01

that's more fun. So if you can be excited about the domain

55:03

and excited about the algorithms, I think that's a

55:06

great place to be. Absolutely agree. All right, we're

55:08

getting short on time, but maybe tell us a

55:10

bit about your data career jumpstart. You referred to

55:12

it a couple of times. Yeah, I have a

55:14

company called Data Career Jumpstart. I just

55:16

try to do a lot of education. So

55:19

the education happens on LinkedIn, happens on YouTube.

55:21

And I actually forgot to mention this at

55:23

the beginning, but I have my own podcast

55:25

called the Data Career Podcast, where I help

55:27

people land their first day to job. We're

55:30

about at 100 episodes. So not quite

55:32

the groundwork that you've put in. That's

55:34

still a ton, that's awesome. Yeah, we're

55:36

getting there. And then yeah, I also

55:38

have a bootcamp where I try to

55:40

affordably help people land their first data

55:42

analyst position by teaching them the skills,

55:44

the networking, and the project and portfolio

55:46

building that they need to do so.

55:49

Like the long version of this show?

55:51

Yeah, basically, yeah, just take what we

55:53

talked about today, expand on it, make

55:55

it like 350 unique lessons, and

55:59

that's exactly. what it is. Yeah,

56:01

very cool. All right. Well, we're about

56:03

out of time. So maybe just every

56:05

final call to action, people maybe are

56:07

inspired, see Diego, have an audience that

56:09

awesome talk very much. So what's next?

56:12

It's easy to be inspired, but

56:14

you got to take action. Yeah, I love

56:16

that. I think it's always fun to listen

56:18

to podcasts, but you probably benefit way more

56:21

from the action you take after a podcast.

56:23

So for you guys who are maybe

56:25

interested in a data analytics or a

56:27

data science career, explore that if you're like, Yes,

56:29

I'm in, you know, make a plan, make a

56:31

roadmap. If you need help, I have like a

56:33

webinar that will help you make a roadmap. Like

56:35

what skills should you learn? How should you be

56:37

networking and stuff like that? But really probably if

56:39

you're just getting started trying to figure out what

56:41

skills you should learn, like what are the top

56:43

skills that you should be learning, and then learning

56:45

those skills, and then not only learning those skills,

56:47

but take action and learning and build some sort

56:49

of a project that we talked about that you

56:51

could put on a portfolio, make a streamlet app

56:53

or something like that. That's probably the best action

56:56

you could possibly take. If you need any

56:58

ideas, advice, feel free to check

57:00

out my website, data career, jumpstart.com or the

57:02

podcast data career podcast. Hopefully there's a lots

57:04

of free resources for you guys to check

57:06

that out. If you've never seen streamlet before,

57:08

I have some YouTube videos about streamlet that

57:10

you guys can check out, but I love

57:12

it. Just take action somehow do something. That's

57:14

one of the huge, huge differentiators is like,

57:16

you might be inspired, but you just got

57:18

to start taking those steps and it becomes

57:20

a snowball. Thanks for sharing

57:22

all your experience and your advice. Hopefully

57:24

some people out there are taking action.

57:27

I'll put out everything we talked about in the show notes,

57:29

of course. Thanks for being your neighbor. Yeah, thank you.

57:31

Thanks for having me. Appreciate it. Bye all. This

57:34

has been another episode of Talk Python to Me.

57:37

Thank you to our sponsors. Be sure to check out

57:39

what they're offering. It really helps support the show. Take

57:42

some stress out of your life. Get

57:45

notified immediately about errors and performance issues

57:47

in your web or mobile applications with

57:49

Sentry. Just visit talkpython.fm

57:51

slash Sentry and

57:53

get started for free and be sure

57:55

to use the promo code talkpython, all one

57:58

word. This episode is...

58:00

sponsored by Posit Connect from

58:02

the makers of Shiny.

58:04

Publish, share, and deploy

58:06

all of your data

58:09

projects that you're creating

58:11

using Python. Streamlet-shinybokeh-fastapi-flask-querto-reports-dashboards and

58:13

APIs. Posit Connect

58:15

supports all of them. Try Posit

58:17

Connect for free by going to

58:20

talkpython.fm slash posit, P-O-S-I-T. Want

58:23

to level up your Python? We have

58:25

one of the largest catalogs of Python

58:27

video courses over at TalkPython. Our content

58:29

ranges from true beginners to deeply advanced

58:31

topics like memory and async. And best

58:33

of all, there's not a subscription in

58:35

sight. Check it out for yourself at

58:38

training.talkpython.fm. Be sure to

58:40

subscribe to the show, open your favorite podcast app,

58:42

and search for Python. We should be right at

58:44

the top. You can also find the

58:46

iTunes feed at slash iTunes, the Google

58:48

Play feed at slash Play, and the

58:51

Direct RSS feed at slash RSS

58:53

on talkpython.fm. We're live

58:55

streaming most of our recordings these days. If

58:57

you want to be part of the show

58:59

and have your comments featured on the air,

59:01

be sure to subscribe to our YouTube channel

59:04

at talkpython.fm slash YouTube. This

59:06

is your host, Michael Kennedy. Thanks so much for listening.

59:08

I really appreciate it. Now get out there and write

59:10

some Python game.