sds podcast episode 351: self-starting in data science · kirill: welcome back to the...
TRANSCRIPT
Kirill: This is episode number 351 with Associate Data
Scientist, Stratos Hadjioannou.
Kirill: Welcome to the SuperDataScience Podcast. My name
is Kirill Eremenko, Data Science Coach and Lifestyle
Entrepreneur. Each week we bring you inspiring
people and ideas to help you build your successful
career in data science. Thanks for being here today,
and now let's make the complex simple.
Kirill: Welcome back to the SuperDataScience Podcast,
everybody. Super pumped to have you on the show.
Question, have you heard Rico Meinl's episode? So
most recently, Rico was on the podcast in episode 335,
and before that he was on episode 123, in January
2018. So today we have a guest, Stratos, who
underwent a really cool transformative journey.
Kirill: He actually heard Rico's episode 123, so the first
original Rico's appearance, and heard Rico's story of
how Rico got on a plane in 2017 and flew to Germany
to attend DataScienceGO, our conference there, and
how that changed his life, and who Rico became after
that. After listening to that episode, Stratos did the
same thing. He's in the UK. He got on a plane from the
UK, flew all the way to San Diego to attend
DataScienceGO, and that also transformed his life.
Now, after pursuing the goal of getting a job in data
science, listening to the podcast, attending
DataScienceGO, doing courses, he'll explain exactly
the courses he's doing, he finally got a job in data
science. So congratulations to Stratos for persevering,
for following his dream. Now he's an associate data
scientist at the National Grid in Warwick, United
Kingdom. How cool is that? Very exciting.
Kirill: In this podcast, you'll learn about how and why he
combined online courses, and which ones, specifically.
We will talk about how to create a data science
ecosystem for yourself and put yourself in that
ecosystem to continue growing and thriving, even if
you're not currently doing data science but you really
want to be. We'll talk about short, mid, and longterm
goals and how to set those for yourself. And we'll talk
about the triad of successful job applications in data
science, something that has worked for Stratos and
surely can work for absolutely anybody applying for
data science, three things to look out for. And as well,
you'll get some interview tips from Stratos.
Kirill: Very exciting episode, very pumped, and on that note,
let's jump into our amazing episode. Without further
ado, I bring to you associate data scientist, Stratos
Hadjioannou.
Kirill: Welcome back to the SuperDataScience Podcast
everybody, super excited to have you on the show.
Today's guest is calling in from the UK. Stratos, how
are you going today?
Stratos: Amazing, Kirill. How are you?
Kirill: I'm doing very well thank you. It's raining today in
Australia. Is it sunny in the UK?
Stratos: No, no. It's never sunny in the UK. I don't know. Let
me just look outside. The sun is just coming up, but I
think it will just be a cold one, but no rain.
Kirill: Yeah, we've got a big time difference, right? It's like
what? Almost 5:00 PM for me, and it's 7:00 AM for
you, right?
Stratos: Yeah. It's just turning 7:00, yeah.
Kirill: That's crazy. All right. What day is it today? Is it
Monday, right?
Stratos: Monday, yes.
Kirill: Monday. Are heading off to work after this?
Stratos: No. I'm working from home today. I already told my
manager that I have a podcast in the morning, but
yeah, usually that's the kind of time I'll go to work. I
tend to work early and finish early.
Kirill: Nice. That's very cool that your job is flexible, with you
being able to work from home.
Stratos: Yeah. I appreciate that for my company, but I think it's
a very common thing in the UK. I think it's something
that we have in the UK that, for my friends in other
places in the world, I don't seeing it being that
common. But in the UK, they're very flexible with
hours and understanding. So, yeah, appreciate that
for the country.
Kirill: Okay. Very cool, very cool. Wow. You have a very nice
and exciting job as a data scientist, congratulations.
That's so exciting, my friend. Well done.
Stratos: Thank you so much. Thank you so much.
Kirill: That is awesome.
Stratos: Yeah. As I said before, the podcast is a huge part. It's
almost like I should congratulate you, as well.
Kirill: That's fantastic. I love how this all unraveled. You
contacted me on LinkedIn, what was this, like a year
ago, in January 2019, right?
Stratos: Mm-hmm (affirmative).
Kirill: You just contacted me to say you're excited, you've
been learning a lot, you've got some data science
opportunities coming up. Then in April, or somewhere
around April, you finally got your first data science
job. I think you started in July. It looks like you
started in July with that job. So that was really cool.
And you just messaged to say, "Hey, thanks a lot." I
really appreciated that. Just looking at it now, I really
appreciate you saying hello and just thank you,
without wanting or needing anything. Of course, that
was a really cool opportunity to bring you on.
Kirill: This is a very exciting success story, of how you went
from not knowing data science at all to now being a
data scientist. So tell us, where did this all start? How
long ago did you decide to start learning data science,
and why?
Stratos: Yeah, so it's interesting because when I started
learning, I didn't start learning data science. I just
accidentally fell into data science.
Kirill: Yeah?
Stratos: I was on my last year at university. That was in 2018,
about March time. So probably two years ago. Yeah,
exactly two years ago. Then I was doing a chemical
engineering degree, and I kind of started having kind
of a passion for programming, or I should call it
automation, or using programming to doing things
better. I played around with a bit of VBA when I was
doing my placement year in PepsiCo. Then I was
reading a few things about Python, that's the language
you should go for. No particular reason. Again, no data
science. So I bought a course on Udemy on Python,
actually from one of your previous guests, Jose
Portilla.
Kirill: Oh yeah.
Stratos: I started getting on with it, and I thought it would take
me about a month or something. Then within a week, I
was almost done by it. I found it so fascinating and
how exciting, you could just manipulate things, data
and all of them. You know how Udemy is with their
recommendations and things like that. They started
recommending some more courses, and some of the
courses were actually your courses. They
recommended to me the A to Z Machine Learning
course. They also recommended to me another course
from Jose as well, which was on data analysis.
Stratos: So I took the data analysis course first because
machine learning stuff sounded a bit too scary for me
at that time. Throughout the course, I started getting
the hang of, "Oh, okay, so you can use pandas with
Python and manipulate data, and you can access
Excel spreadsheet. You no longer have Excel
spreadsheets. You have this data frame structure,
where you can do whatever you want," and all of this.
Then I started going into the space of visualization and
using Matplotlib and seaborn.
Stratos: I don't know how much detail you want me to go to,
but what I decided then is I actually like this space. So
I kind of took a step back, did a bit of research on
what areas you need to learn. I then started doing
something that I think worked quite well. I started
combining courses, instead of just doing ... I used the
method that you use in university, how you would
probably have lots of courses simultaneously. It's not
like you have one course and then next one.
Stratos: What I did, I took your course, the A to Z Machine
Learning, because I decided I wanted to learn machine
learning. I took the A to Z Machine Learning course,
which was very practical, full of examples, amazing
intuition, videos by yourself. Then, because I had that
extra math knowledge from my background in
engineering, I also wanted to dig a bit further. In
combination, and you might not know that, your
course maps very nicely with Andrew Ng's course, from
Coursera.
Kirill: Which one?
Stratos: The machine learning course, the very well known
machine learning course. What I mean they map very
nicely is, you guys start with linear regression, if I
remember. He starts with linear regression. You
switched to ... So it's almost like you go how I did it,
but obviously it's up to the listeners to do whatever
they want. I would start with Andrew Ng's, which is
very technical, very mathematical, but maybe lacking
that application. Then come to your course, listen to
your intuition videos, kind of confirm that, yep, I
understand it. Then bam, go into the practical. That
kind of got the ball rolling.
Stratos: Within a month, I covered all of your course and
Andrew Ng's course. Yeah, and then continued on. I
continued the same logic with deep learning because
Andrew Ng had also a deep learning course and
mapped it to your Deep Learning A to Z. So it worked
quite nicely. I didn't know if you knew that about your
course, but your course is-
Kirill: No.
Stratos: My understanding is you wanted to create a course
that could be intuition based, people will kind of get on
hands on, but it also works for people are also
interested in the mathematics. I feel like your intuition
videos are very nice to confirm your knowledge,
without needing to dig further and derive those
equations again, if that makes sense.
Kirill: Mm-hmm (affirmative), yeah.
Stratos: Yeah, so I found that very fascinating.
Kirill: That's very cool. Sounds like we need to partner up
with Andrew Ng and create a course together. That
would be cool.
Stratos: Yeah, there is also some nice books that I was reading.
I think it was the Hands On Machine Learning. I can
find exactly the title for it. Again, that book starts off
with the basics of machine learning, linear regression,
all of these, and then it proceeds into Tensorflow and
deep learning. You can have that book on the side
while ... If you don't understand something from the
two courses, or if you want to learn a bit more, then
you always have the book to reference.
Kirill: Gotcha.
Stratos: Yeah, that was quite a nice experience. I learned quite
a lot from that.
Kirill: Interesting. Tell me this, because those courses are
massive. For instance, the Machine Learning A to Z
course is 40 hours long, right? It's huge.
Stratos: Mm-hmm (affirmative).
Kirill: So what I'm wondering is, how do you keep up the
motivation and also ... I don't know. How do you
supplement that aspect that you don't have a full-time
data science job where you would apply these things?
So you're learning, that's great, but then you go to
your work and you're an engineer. You're doing
something completely different. So how do you keep
that ball rolling? How do you keep the momentum?
Where do you get those hands on applications, to keep
you excited, to show you how you can actually apply
this knowledge in the real world?
Stratos: Yeah, that's an excellent question. I have to make a
disclaimer that I started learning before I actually
started the job. It obviously continued while I was
working as an engineer, but my main kind of biggest
learning was during summer holidays. To your point-
Kirill: So you started before you even started your
engineering job?
Stratos: Yeah. So I graduated in June. Then around July was
about when I started doing all of this that I just
described. I was starting a new job in September, so it
kind of followed through all the way until I got a job in
data science.
Kirill: Gotha.
Stratos: The first thing I did, which you might be thinking that
I'm copying Rico here, but I think it's your machine
learning course that has ... One of your courses has
Rico's podcast on one of the notes in Udemy. So I
watched that. Actually, that's how I found out about
the podcast. I went back and re-watched the previous
episodes, and he-
Kirill: How many episodes have you listened on the podcast
to?
Stratos: Oh, I'm up to date now.
Kirill: All of them?
Stratos: Yeah, yeah. I think I haven't listened to the 341, the
one that just came out, basically.
Kirill: No way! That's amazing, my friend. That is crazy.
Stratos: Yeah.
Kirill: You've listened to 340 episodes?
Stratos: Yeah. No, the podcast is ... Actually, yeah, to your
question, that's one way to keep going, is the podcast.
Having something, especially the podcast because it
comes every week. It's almost like you've got seven
days. Yeah, you're bound to lose motivation on some of
them, but at the least, every seven days you've got that
motivation to get you back. On the podcast, you hear
people progressing, so you're like, "I can't stay where I
am. I need to progress as well."
Kirill: That's true, yeah.
Stratos: So the podcast was a big one. But yeah, coming back
to Rico's podcast. I heard about the DataScienceGO. I
just went on the website. I saw it. I was debating,
should I go? Should I not? Is this for me? I'll be
honest, it looked like it wasn't for me, in the sense
that, who am I? I'm not a data scientist. I only just
started learning a few things. I don't even know what
machine learning means. But then I decided, "You
know what? If Rico did it ..."
Stratos: I decided to just book a ticket and go to the US for two
days and come to DataScienceGO, which looking back
now, it was probably the best choice I could have
done. I'm being honest here, it's not that because I
came to the conference I now have a job. It's not that,
but it's the mentality. It's almost like that milestone:
okay, if in October I need to be at this conference, then
until October I need to upscale myself. So I don't have
any time to lose. Then obviously after the podcast,
with all that pumping that you leave the podcast,
you're kind of ... Yeah. So I think that was a big thing
that kept me going. What do they call it? What did
Rico call it in the second phase?
Kirill: Radical commitment.
Stratos: Yeah, something like that. Basically putting something
in the diary that you know you just can't miss. You
paid for the ticket. You paid for the airplane ticket. I
mean, unless you're somehow insanely rich and you
don't care about money, you might as well just go. So
that was one.
Kirill: Let me clarify this. So what Rico did, and he's been on
the podcast. Actually, he was on the podcast recently,
again.
Stratos: Yeah. I listened.
Kirill: When was that? He was again on episode 335, but the
first time was a year before that. Or maybe more than
a year. But basically what Rico did, a crazy thing, he,
from Germany, booked a ticket to come to
DataScienceGO in the US, in San Diego. I think this
was the first one in 2017. Just for that. He just flew
there, came to the event, and then flew back. That
completely changed his life. So you're saying you did
the same thing but in 2018, right?
Stratos: Yeah, and from the UK, not Germany.
Kirill: From the UK. That is so cool. Did you get to meet
Rico?
Stratos: I did. Well, not in person directly, but I did speak to
him after the ... Yeah, you can say yeah I did. I spoke
to him briefly after he spoke, and I also asked a few
questions. But no, sadly I-
Kirill: Was it cool? Was it cool to see that person that
inspired you to fly across the Atlantic, to see him in
person?
Stratos: Yeah. I think what was more inspiring was his talk
during the podcast.
Kirill: Oh yeah.
Stratos: Sorry, during the conference.
Kirill: During the conference?
Stratos: Yeah, it's the fact that he was there. It's almost like,
thinking through the timeline, that a year ago he was
just this student that flew here, and then he was
there, standing with all that confidence and spreading
the word out there. Yeah, it was very fascinating, to
see him out there. Congratulations to him for
everything he's doing.
Kirill: That's awesome. Well, coming to the event or events is
another way to supplement your learning. Listening to
the podcast, coming to events, and all those things
together keep you going. It's very inspiring to hear that
you were able to create this kind of ecosystem for
yourself, you know? I don't know how many people
around you were studying data science as well, but
you tapped into the SuperDataScience community, the
DataScienceGO community, and by doing that, you
kept yourself propelled and motivated to go forward. Is
that about right? Is that how you see it?
Stratos: Yeah. I think, yeah. I've never heard of it this way, but
yeah. Putting it into an ecosystem, yeah. That's
correct. It's very important, I think, to keep having
some sort of a short-term, medium, and longterm
goals. You need to know why you're learning what
you're learning. You can't just learn for no reason. I
mean, yeah okay, you can learn for your benefit. I get
that. But you need to be aiming for something. "Oh, I
want to learn this because I think this will help me
reach that goal." If that makes sense. That would be
my advice to everyone, to always have a goal, always
know why you're doing something.
Kirill: Okay, so tell us about your goals. What were your
short, medium, and longterm goals when you were
learning data science?
Stratos: The short-term is I just knew that any kind of job that
I will be doing, engineering or not engineering, I knew
that I needed to know programming. I can't just stay
with Excel or whatever, with Word and all this. I need
to be able to do things faster, do things better. I think
that knowing programing is something that everyone
should pursue. I'm not saying everyone should be an
expert, but having the ability to program, doing things
faster, automating, it just takes the boring aspect out
of your job and makes everything more interesting. So
that was kind of the short-term goal.
Stratos: The medium and the longterm, kind of it's a bit of both
together. It started off I just want to influence data
science everywhere, but then after DataScienceGO,
when I came back from the US the first time I went to
DataScienceGO, I said, "Yeah, okay, now we're shifting
goals." My medium/longterm, because I didn't know
how long it would take, I decided I want to be a data
scientist. It was no longer a hobby for me. I'm
spending a lot of time, of my own time, afternoons,
nights, doing that. I might as well do it for a job, and I
might as well be actually practicing it properly and in
a place where my skills could be of maximum help. So
that was kind of my goal.
Stratos: From there, as soon as I kind of put that goal into the
books, that's it. My learning became a lot more
consistent, a lot more structured. I was looking on
what is asked in the market, and I was learning it.
That kind of got me going. From there, I had no kind of
throwbacks. I wasn't going back.
Kirill: Okay, gotcha. So short term goal, learn programming
because you're going to need it anyway. Midterm goal
was you love it so much, you might as well get a job in
a it. What was the longterm goal?
Stratos: Well, the longterm goal, and I don't know how to define
it, but whatever I'm doing, I want to be influencing
data science. I like doing data science. I like producing
nice charts, models, and things like that. But what I
particularly get hyped about is when I show people
who don't know about data science, don't know about
programming, what it can do for them. I like the idea
of going somewhere and kind of disrupting that
organization or that team or whatever, from the
concept of data science. "You know what? You no
longer need those Excel spreadsheets. You can display
them in tableau. You can run a model that gives you
predictions. You can save so much time." That's my
longterm goal. Whatever I will be doing in I don't know
how long, I want to be influencing data science. I want
to be the person that will go in, and after I left, data
science has exploded. I don't know if I'm making it too
generic, but that's kind of my longterm goal.
Kirill: Okay. So you want to be spreading data science.
Stratos: Yes.
Kirill: Across different companies. Gotcha. Very exciting.
That shows a lot of passion, that goal. That's
something very admirable, admirable to have that kind
of goal. Very cool.
Kirill: Take us a bit back. You did all this learning. You went
to DataScienceGO. How and when did you start
applying for data science jobs?
Stratos: When I came back. Do you remember ... I mean, I'm
assuming you do remember. At the end of the
DataScienceGO, you gave us these-
Kirill: Which one? 2018 or-
Stratos: 2018. You gave us this kind of talk, where you said-
Kirill: Yeah, yeah, yeah.
Stratos: Close your eyes and right something down, or
whatever. Anyway, I remember I had this small
notebook that you were handing out in the conference.
When I was on the plane, 10 hour flight, which-
Kirill: For context, the exercise was we needed to get up, we
needed to make the sound of victory or something like
that, feel really empowered and passionate. Then
imagine success, what you want to accomplish in the
next 12 months. Then the objective was to sit down
and write down your top three goals, right? Was it top
three or top one?
Stratos: I mean, I wrote only one, but it might have been top
three.
Kirill: No, top one.
Stratos: It might be top three.
Kirill: Okay, so it was top one action you're going to take
when you get home. That was the thing. So you're on
the plane. Sorry, let's get back to your story. You were
on the plane.
Stratos: Yeah, so from that moment, when you said imagine
success, I might sound cheesy, but kind of I felt it
there that my success, for me, at least within the next
12 months, was me standing somewhere, anywhere,
and being a data scientist. Because I wasn't at that
point. I wasn't applying. I was just a self-learner. So
that was kind of locked in as a success.
Stratos: Then on the plane, I just started thinking of what can I
do? Because I was clueless. I just knew I'm about to
leave this very kind of prestigious engineering firm and
start going into probably one of the most competitive
fields, having an engineering degree, which let's face it,
is not ... I mean, you hear mathematics degree,
physics degrees, PhDs. It's a good degree, but it's not
the best.
Stratos: So my first plan was, okay, let's see how other people
do it. It was evident to me that the way to get out there
is to physically start shouting about yourself. The first
thing I decided to do on the plane was I-
Kirill: Shout.
Stratos: Yeah, shout in the plane and see if anyone is hiring.
No, it was just literally what's the best way for me to
show what I'm doing? As you can imagine, the first
thing that came to mind was LinkedIn. I'm not a
particularly social media person. I probably haven't
changed my profile picture for five years. I just have
them, nothing more. But I decided let me get out of my
comfort zone and start posting. That was one thing I
decided to do.
Stratos: The other thing is, as I said before, I need to start from
somewhere. I can't just go nuts, start updating my CV.
I need to get very specific because I don't have a lot of
knowledge to kind of brag about. So I went back and
did a bit of ... Sorry, again, I'm still on the plane. I said
I will sit down, read what's out there in the market.
That's another thing, I didn't know what's in the
market. Because coming from DataScienceGO, all I
would hear is Silicon Valley, Silicon Valley. I'm like,
"Yeah, I'm very far from Silicon Valley."
Stratos: So I need to see what's in the UK market. That was the
first thing. More specifically, see what they're asking
for, what skills are out there, and obviously how do I
compare to those. From that, start developing the
corresponding skills so I can at least put them down
on the CV, so when that first scan goes through ...
Because I was feeling like if I get to the interview, if I
kind of meet their technical requirements, I feel that
story and that passion, I would be able to get out there
and hopefully it would be enough.
Stratos: Then the final thing I said, because I think that came
from Ben Taylor's talk, I need to get myself some
applications. I can't just say, "Oh, I took Carol's
course," or, "I took that course." I need to get myself
some toy applications, even if they seem toy examples
and things that nobody cares about. They need to be
out there, so I can demonstrate that I don't just know
how to follow a course; I actually know some
applications. So that was the three main goals.
Kirill: Like projects you mean, right?
Stratos: Yeah, yeah.
Kirill: Like a portfolio project.
Stratos: Yeah, portfolio or just find ... To be fair, what I took
from Ben Taylor's talk was it's not just making a
project. It's just, okay, you know data science; find
something that you're curious about and just have a
play with it. That was my approach.
Kirill: Yeah, that's really cool. That's very cool. An example of
something like that, a recent one I heard, I was
listening to this one video, just briefly, of I think the
CEO of Kaggle. They were describing how they had this
one competition with some data sets there, and it was
about people doodling, like drawing things, drawing
animals or whatever, and the algorithm had to detect
what animal. Was it a lion? Was a hippo or an
elephant? Whatever else they were drawing. So people
would not only do that, but they would go an extra
step and they would try to understand, depending on
your cultural background, are you more likely to draw
the animal clockwise or counterclockwise? How is that
distributed by country? Crazy stuff, right?
Stratos: Yeah. That sounds-
Kirill: Like you say, whatever your passionate about, use
data science and come up with some insights.
Stratos: I'm a big fan of Medium, and you hear some people
just putting some random dataset and some random
projects out there. You're like, the only way you could
have come up ... The most recent one I've seen, and I
was like, whoa, that's fascinating ... Are you familiar
with the application Tinder?
Kirill: Yeah, of course.
Stratos: Okay, I'm not that familiar. I haven't used it that
much. But this guy had a tremendous amount of data
from his Tinder account, of his friend's Tinder
account. So he asked-
Kirill: "His friend's Tinder account," in quotations marks of
course.
Stratos: Yes, yes. Yeah, it is not him, yeah. And it was amazing.
If I remember correctly how the application works. He
did this diagram of how he started, with how many
likes he did, how many super likes, how many of them
ended up being liked back, how many ended up in
conversations, how many of them replied, no reply.
Basically, towards the end of the chain, how many
actual successful dates and something like that. Some
crazy metrics. It just shows the application of data
science.
Kirill: Wow.
Stratos: As long as you have some data, you can just get some
insane [crosstalk 00:30:22] out. Yeah.
Kirill: Interesting.
Stratos: I bet you, actually, if that guy was a publishing a
dashboard or something, Tinder would have picked it
up and put it on their application as a dashboard.
Wouldn't you want to see what's going on, to
automatically have your data?
Kirill: Yeah, yeah. Yeah, there you go. If he wants a job as a
data scientist at Tinder, he's got it, right?
Stratos: Yeah.
Kirill: He just needs to send them this link, and they will be
like, "Oh my god, he loves our product. He loves data
science. What else can we want? We got to hire him."
Stratos: Yeah, that's the thing. He doesn't even need to send
his CV.
Kirill: Exactly. It's interesting, actually we are hiring for ...
What are we hiring for? For like a product coordinator
at SuperDataScience. I was reviewing these three
applicants recently, and three of them ... This story,
it's not about data science applications of course, but
it's still relevant. So these three people ... A lot of
people applied. A lot. We got, I wouldn't say millions,
but we got quite a few applications. Then in the final
round, there's three people. I get these three CVs.
Actually three profiles of people, like three emails, one
about each person.
Kirill: So I read the first one, the second one, the third one.
Then for the third one, I'm reviewing his profile, and I
noticed ... I'm reading his CV, and I'm like, "Hold on, I
read this CV for the second person, for the third
person, but I don't remember seeing the CV for the
first person." So I went back to look at the first person,
the email I got about that person.
Kirill: I realized that there is no CV. There is just I think
there was their LinkedIn and a website that they put
together, that describes them, like what they're
capable of, what kind of designs they've done, what
products they've created, and things like that. So it's
kind of, like you said, a portfolio project. I never even
got a CV. That person doesn't even have a CV, In the
end, that turned out to be the best applicant, and we
ended up hiring them. You don't need a CV these days.
You just need to demonstrate that you can do things.
Stratos: Yeah, and do things differently, I think. That's what I
get out of that story, is if you are kind of innovative
enough to show exactly the same thing but in a
different way, that's what will stick out to the
recruiter.
Kirill: Exactly. That's a really good point. So let's recap,
before we get too far away. Let's recap on your I'm
going to call this the triad of successful interviews,
right?
Stratos: Yes.
Kirill: What was the first, second, and third items?
Stratos: The first one was ... I forgot. The first one was-
Kirill: LinkedIn.
Stratos: Yeah, make yourself-
Kirill: Have a LinkedIn.
Stratos: If you want to generalize it more, make yourself visible,
that you are doing what you are doing. The second one
was I think ... Was this the second one? Anyway,
basically search the market. Make sure, know what
jobs are out there, know what you want to get, and
ultimately what they are asking for, so you can develop
in those areas. The final one was don't just limit your
... Given that you're just a self-learner, don't limit
yourself to just courses. Start putting some
applications together.
Stratos: Let me just say that it's not just about ... Obviously
what we just talked about, yeah, it's beneficial for your
[inaudible 00:33:46]. But it's not just that. It's also like
a self-confirmation, that you actually like this space.
For me, if you can spend your Saturday looking
through some data from Tinder, let's say, instead of
doing something else.
Kirill: It doesn't have to be Tinder. It could be Airbnb. It
could be Uber rides. It could be your recent google
searches. Whatever. There's so many data-driven ...
Like your Netflix movies that you watched. Whatever
comes to mind.
Stratos: Exactly, yeah. If you are willing to spend your nights,
your afternoons, or whenever, whenever is a free time
for you, to do that, just for fun, then it's almost like a
self-confirmation. Yeah, that's where you need to be.
Kirill: Okay, gotcha. Great. So the triad is LinkedIn/make
yourself visible, know what jobs are out there and
what are the requirements, and number three, have a
portfolio of projects that will talk for you.
Stratos: Mm-hmm (affirmative).
Kirill: Okay. All right. Very cool. So you made yourself
visible. Eventually did you find the jobs you wanted to
apply for? Did they find you? How did you go from
there?
Stratos: Because I already had a job, I was a bit picky. I didn't
want to just go crazy and start shipping CVs all over
the place. So I became very, very picky on the
applications. Yeah, I did find a few applications that
looked like ... I wanted something that looked entry
level, but also had not just entry level but also some
development opportunity. Because I knew that if you
were to put me in the data science spectrum, I need at
least six months for me to understand how the
industry works, to fill out those gaps that you develop
as you go through the self-learning experience. So I did
end up finding a few opportunities like that, mainly
through LinkedIn.
Kirill: And you applied for them?
Stratos: Yes, yeah.
Kirill: Okay, and so how many did you apply for? How many
did you hear back from?
Stratos: I think I applied to five. I know some of your listeners
will be like, "Just five?"
Kirill: Just five?
Stratos: Who is this crazy person? Yeah. But I spent so much
time in those applications that it almost felt like five
each. So you can take that as 25. I think I heard back
... The one I got immediately the job, but the other four
I progressed to the interview stage.
Kirill: Wow.
Stratos: Yeah. Two of them I did the interview ... No, let me
start with the easiest one. Two of them I reached all
the way and got offered. Which, one I took. The other
two, I think I reached both of them, the interview. One,
I got rejected at the interview. The other one, they were
just not ... I did the interview, no response. I already
got the job here, where I am at the moment. I couldn't
be bothered, so I just ... I don't even know. Maybe they
replied at some point, but yeah. For me, if someone
tells me I'm going to come back at you in two weeks,
okay, if it's three weeks that's fine, but if you tell me
two or three weeks and then they don't come back for
like three months? It kind of puts me down. Why
would I want to work for an organization that doesn't
even bother telling me, "You know what? We can't
come back to you at the moment. We need some more
extension." So yeah, I just left it.
Kirill: Okay. Very cool. So you applied for five. Four of them,
you got interviews. That's an 80% success rate to
getting interviews.
Stratos: Yeah.
Kirill: Then two out of the five, you got offers. Meaning, that's
40% success rate getting a job offer. That's crazy, man.
Congrats. That's awesome.
Stratos: Thank you so much.
Kirill: Really exciting. To anybody who says, "You only
applied to five? I know people who have applied to
hundreds of jobs." Well, the success ratio there is like
0.001. It's much better to have a high success ratio
and know that you're applying for jobs that you really
actually want yourself, that you're passionate about.
Yeah, it's better to spend more time on one application
and tailor it and really understand the company,
understand their mission, understand how you can
help, and have that laser-specific conversation with
them. Rather than just sending this template email to
hundreds of companies and hoping something ...
What's it called in shooting? Spray and pray. You
shoot all these applications, and you just hope and
pray and wish that, "Oh well, hopefully somebody will
reply, and then I'll take that job." You'll end up in a job
that you don't love, anyway.
Stratos: Exactly, yeah. It pays off. It's very tempting, when you
are either desperate for a job or when you want a job,
to just quickly update your CV so it looks generically
okay and just spam it. Especially nowadays with
LinkedIn, when some of them are just easy apply. You
just click a button. Done. You applied. It could be your
next job. So I think it's very tempting to do that. I'm
guilty myself. I've done it in the past, not for data
science, when I was applying for engineering jobs. But
it's as you said, it comes down to learning the
company, understanding whether it's where you want
to work for. Even if you do get the interview, if you
didn't spend that much time learning about the
company, you're very likely not to be successful
because you will come off as just a random person
who sent their CV, just because of the salary figure or
because they just know the company because they're a
well-known company. I think it all pays off.
Kirill: No, totally agree. Okay. So cool. We probably won't go
too much into the interview process. Actually, yeah,
let's talk a bit about it. Was there a lot of technical
questions on the interviews? Was it more behavioral?
Is there anything you can share? Any tips you can
share for people listening?
Stratos: I had, as I said, four interviews. All of them were very
different. It goes to show how different [inaudible
00:39:54]. So I will just talk kind of generically how.
One thing that is very common, you will think, "Oh, I
need to know Python, or this." Yeah, you need to know
you are likely going to be asked to do some exercise. In
one of the interviews, I was asked to do an SQL
exercise, and that was on the spot. In another
interview, I was sent some data three or four days
beforehand, and I was asked to do some analysis, just
go back and present my analysis to them. In another
one, it was just like either do this exercise or just bring
something to talk through. So whatever you're
applying for, expect some sort of an application.
Stratos: One tip I would say to that is just because you know
Python and you put down Python, especially if they are
asking for more than Python, let's say they are asking
for R, or they are asking for Java, whatever, it depends
which company, but don't be surprised if you go sit
down, you're amazing in Python, and they give you an
introduction to R script to run through. It's just for
them probably to test whether you will be dealing okay
with other languages.
Stratos: If you want a real-life scenario, I am known to know
Python. All of my team are known to know Python, but
because we have a historic model that was written in
R, and now I'm working in that model, I now have to
write in R. I've never touched R before. But now I have
to learn it, and we have timelines. That's just how life
works. We can't just change the model to Python
because it's what we know. So that's one kind of thing.
I know it sounds terrifying, but just be prepared for it.
Stratos: It happened to me, thankfully not with R. It happened
with SQL. I knew a bit of SQL. When I sat down, I was
expecting for them to ask me some very technical
Python questions, and they just were like, "No, we just
want you to connect to the database, query a few
results, and do some group [inaudible 00:42:04]." Now
that I know a lot of SQL, it's like that's nothing. But
back then, whoa, okay. So it's kind of a bit of a
terrifying thing.
Stratos: More importantly, outside of the technical things, do
expect to get some soft skill questions, specifically
related to data science. That can be in a direct
question, such as, "How would you deliver the results
of a model to the execs?" Or, "If someone was going to
give you an Excel spreadsheet to present something,
what would be your best approach? Would it be a
visual? A model? Which one would you use?" Some
other questions you might get. Someone might
describe their problem. Someone might come and say,
"I have this kind of assets, and they're all failing. What
do I do?" That's where they will ask you to kind of
formulate the problem and kind of decide on how to
approach it.
Stratos: Or, it could just be a presentation. That's what was
going on in one of my interviews. I was asked to
present my code, and then I was given 10 minutes to
present a presentation that was targeted to non-
technical people, so someone who doesn't know code,
someone who just wants to know what's going on.
Those skills are very important. That's what people are
looking. At the end of the day, you can train anyone to
become an expert in Python, but you need to have that
ability to talk to stakeholders, pass on the message,
formulate the problem. Those are the important
things, at least for me.
Kirill: Wow, fantastic. I'm really glad that companies are
testing that, now. Back in the day, when I was
interviewing, it wasn't a big consideration. But I think
more and more companies are realizing that these soft
skills are important in data science, at least as
important as the technical. Because if you can crunch
numbers and get the insights but you can't
communicate them, then what's the point in that?
Stratos: Yeah. I think one thing that you will like, my manager
says that a lot, but it's the 80/20 rule. I think you had
it in one of your podcasts, as well. That's a very hard
thing. What I mean by the 80/20 rule, for those who
are not familiar. I mean, if you've got a piece of
analysis, usually to reach to the first 80% of getting
your message across, it would take you let's say a day,
or two days, if you do small time analysis. To get that
extra 20% and make it amazing, you probably need
two or three weeks. So getting the ratio right is
incredibly hard, incredibly hard. If you want to do well
in business, or in the industry of data science, you
probably need to become an expert in that 80.
Stratos: The golden rule is get that 80, minimum viable
product, as soon as you can. Get it to the customer,
and if they want that extra 20, which most of the time
they don't, then put the time. Don't put the time
beforehand, because usually that 20 will be wrong,
unless you ask your customer beforehand. I don't
know if that makes sense, but that's a very, very key
principle.
Kirill: Exactly.
Stratos: And I learned it the hard way.
Kirill: What do you mean? Like on the interview?
Stratos: No, when I started my job.
Kirill: Ah, okay.
Stratos: On my first problem, it was a very relatively easy
problem, but I wanted to impress everyone. I wanted to
do amazing. I found myself spending two weeks on a
problem that should have taken me a day. That's
where I kind of was introduced to that rule. Since
then, I kind of go by it.
Kirill: Fantastic. Wow. Speaking of starting, of first problems,
can you share a bit about that? What was it like when
you started your first job in data science? Was it how
you expected it to be, or was it completely different?
Stratos: The first time, coming from self-learner, the first kind
of month or second, it was just getting my head
around that I write Python for a living. That was just a
very nice feeling. Because you're used to coming back
from work, I have to do a bit of courses. It's almost like
you're hiding away writing away your Python, and
you're feeling like you're doing something illegal. Now
you're legal. You're allowed to write Python and get
paid for it. That was an amazing feeling.
Stratos: What was very good, and I really, really appreciate my
manager for this, is when we started, the first thing we
did, we sat down. And I encourage every manager, or
everyone who is going to start coaching people to do
that. We sat down and figured out, mainly based on
my feedback from the interview, we sat down and
discussed what are my weaknesses and what do I feel I
need to develop at. Mainly technical, as in, "You need
to develop your statistics. You need to develop your
Python, your version control."
Stratos: Once you do that, you do that first week or something,
then you can tailor your projects around them. You
can focus a bit more. You can ask for work that is
more focused in those areas. What I'm trying to say is,
when I started, I was a bit lost, but once we had that
chat and I knew the areas that I had to develop, kind
of that stress went away. I was like, "Okay, here's what
I need to know." It was almost like I was coming back
to my self-learning days, but now I was learning for a
bigger purpose. I was learning it to apply to work.
Kirill: Wow, fantastic. Probably experiences in different
companies are going to differ, but it's still really great
to hear how yours was. I think before the podcast you
mentioned that the first three months or so were all
learning for you, before you could actually start feeling
that you're fully working.
Stratos: Yes.
Kirill: Tell us a bit about that. Was that a frustrating
experience? Or was that fun, that you were actually
learning for work?
Stratos: It was a bit of both. It was fun when it was working. It
wasn't fun when it was not. I think what I would
struggle in is me not being from a mathematics or a
physics ... Well, let's stick to mathematics and
statistics background. Sometimes I was getting
frustrated with myself for getting the basics wrong. I'm
talking very basic, like confidence intervals and even
means and medians and things like that. I was just
getting frustrated with myself for not getting it. So that
was a bit annoying, but again, because you've gotten
the job, you do the mistake once, you do the mistake
twice, and you learn. The first thing I did, for example,
is I went back and freshened up my statistics: how will
you calculate confidence intervals, how you would you
do hypothesis tests. So then I could kind of not make
those mistakes again.
Stratos: Then in terms of the soft skills, it's very hard going
from staring at a laptop and hearing someone teaching
you things and doing data science, to going out in the
real world. I literally thought that ... Okay, I'm not so
delusional, I didn't think that, but you want to think
that you'll go into the industry, you get a meeting
invite, like, "Hey, I heard you were a data scientist.
Could we have a catch up?" You sit down with the
customer or the stakeholder ... By stakeholder, sorry,
that's what we call them, I mean internal people. I'm
part of the company. I'm not external at all. So
someone will invite me, we'll sit down, and I thought
they'd be like, "I have this data. It looks amazing. It's
ready for you to split it and fit into the model. I just
want the predictions." Basically what you get in a
course, where you get a nice data set. Maybe it's a bit
dirty, but you need to clean it.
Stratos: I didn't really get the fact that to even reach the point
to actually have data, it's like a marathon. That was
very frustrating in the beginning. Now, it's very
rewarding because, again, I'm coming back to the
marathon, but if you are very good at formulating the
problem and making sure that you know what your
customer wants, that marathon gets shortened and
shortened and shortened. You go from long and long
discussions and confusions to, "Okay, I can hear what
you're saying. I know what your problem is, and I
know how to solve it. Do you have this data set? Or are
there any ..."
Stratos: That art I will call it, that art, knowing how to
formulate a problem, how to go from a problem to data
to solution, it's very frustrating in the beginning. I
mentioned it before the call. It's very hard to practice
as a self-learner. Now, to be honest, probably very
hard to practice even if you come top universities and
data science courses. I think that's the on the job
training. It's very hard to practice.
Kirill: Yeah, fantastic. No, I completely agree.
Stratos: Lots of frustration there, but lots of reward after you
kind of ... I'm not saying I mastered it, but now I feel
like I understand it a little more.
Kirill: Yeah, and you learn to appreciate it, as well.
Stratos: Exactly, yeah.
Kirill: Fantastic. Well, that is awesome. Thank you for such a
great description. We're running out of time on the
podcast, but we could keep talking for ages about all
this stuff. You've had such a really cool journey,
Stratos. I'm very inspired to hear from you and very
excited that you came on the show to share with your
colleagues and friends and fellow data scientists. I
think this is going to be, like you said at the start, a
push for others to keep going. Once they hear your
story, they'll be thinking to themselves, "Well, I can't
stop now. I got to keep going. I got to move forward."
Stratos: Yeah.
Kirill: Very cool. Very cool. Well, before I let you go, please
could you share with us where are the best places for
people to follow you and your career?
Stratos: I think the best way is probably LinkedIn, because
that's the only place that I stay relatively active then. I
do encourage, if someone has any questions on how to
get started and where to go from where they are, or
any suggestions, I'm more than happy to help people.
It kind of ties very nicely with my longterm goal I said
before, where if I can influence data science, I'm more
than happy to do it. So yeah, I'm happy for people to
get in touch. I can answer any questions. Or even if
people are around from the UK area, and they want to
catch up in person. I'm more than happy.
Kirill: What city do you live in, in the UK?
Stratos: In Warwick. I don't know if you know Warwick.
Kirill: Warwick. We have a Warwick in Australia, as well.
Stratos: It's probably a lot bigger than Warwick. Warwick is a
town. It's not even a town. It's a village.
Kirill: Oh, gotcha.
Stratos: But yeah. It's close to Birmingham, if you know where
Birmingham is.
Kirill: No, I don't, but I'm sure people [crosstalk 00:53:31].
Stratos: Yeah. If someone is from the UK, I'm sure they will
know Warwick.
Kirill: Yeah. Stratos, I got to make a comment for you. I'm
looking at your LinkedIn, and it looks like last time
you posted was four and five months ago, and then
before that it was a year ago.
Stratos: Yeah.
Kirill: Looks like somebody got a job in data science and
stopped posting.
Stratos: No, that's actually something I wanted to talk about.
You know how I mentioned that I wanted to get out of
my comfort zone and do ... It's fine to get out of your
comfort zone, but unless it's something that you
actually enjoy yourself, then ... The reason I stopped
posting is not because I got a job. It's because I found
it very stressful, in the way that I didn't feel like I was
getting much of it. I very much prefer to have someone
come to me and we talk and I help them personally,
rather than me trying to post. It's just not me. I mean,
I don't know how to describe it. It's not myself. It feels
like I'm portraying a person that's not me, if that
makes sense.
Kirill: Okay, well-
Stratos: I do post when I find something interesting, but yeah.
Kirill: Well, I have a piece of advice for you, then, if you don't
mind.
Stratos: Yeah.
Kirill: You shared a bit of advice. Can I give you some, as
well?
Stratos: Yeah, of course.
Kirill: Coming and talking is amazing. That's a fantastic
thing. At the same time, being not yourself is terrible,
right? You want to be yourself. You want to be not
necessarily comfortable, but you want to feel like
you're doing something that you enjoy, or you can
enjoy with time.
Stratos: Exactly.
Kirill: However, people are not going to come and talk with
you, or won't be able to find you, or even know that
they can talk to you, unless you somehow get on their
radar. So my advice would be, I can totally understand
if posting is not your thing, find what is your thing.
There's so much medium out there online. You can
post. You can write articles. You can comment. You
can reply on Quora, answering questions. That's not
posting updates. It's answering questions. Ben Taylor
was the number one AI influencer on Quora because
he answered a lot of questions. You can record videos
for YouTube. You can record audios and share them
on SoundCloud. You can become a mentor on one of
these mentorship platforms.
Kirill: Basically, find something. Not even posting. For
instance, you could do projects. I'm sure you not only
do projects for work, which are sensitive, but probably
you still do projects for your own portfolio, or maybe
some here and there you'll do a data science project
where you can desensitize things, and you can post
those. Or you can publish them on GitHub or on
Kaggle or on Tableau Public. Even without writing
anything, you can publish things. Like I have a profile
on Tableau Public, which I haven't updated in a long
time. I used it while I was creating the courses in
2015. I looked at it, and it's got thousands of followers
because I shared useful dashboards with people that I
didn't even have to write anything. People look at
them. They can click on them. They can download
them. People like them.
Kirill: My point is, I completely appreciate that maybe a
certain type of medium is not yours and you don't feel
yourself, but find what is yours and do that. Because if
you want to have those amazing conversations and
meet your mentors and meet people who you can
influence and spread the word about data science,
they have to be able to find you somehow. It's really
hard. This podcast is going to help for sure, but don't
stop. Find other ways that you can help. Maybe start a
podcast of your own. You never know.
Stratos: Yeah, and I appreciate that. Actually, on that, it's
something that I've been thinking through. You are
right, and I think what I want to get out there is, yeah,
one of the things you said is doing projects. I do do
quite a lot of things outside of work, more projects.
That's kind of my thing. I'm the kind of person which,
if I see a data set online, I'm curious what that looks
like in a graph. I will just plot it. I do quite a lot, and I
think I need to think through what's the best way. I'm
happy for the listeners to suggest what would be the
best for them, but if there is any way for me to kind of
get that going.
Stratos: I think that's my strength, is I can help people and
advise people how can they get that step and go and
get a job. I can probably not advise them what's your
best model to use. I'm not that technical. But I can get
you from I don't know if I like data science to I'm
passionate about it. So that's the kind of area I want to
focus on, my community if you like. That's the kind of
community I want to target. But in what way and what
medium, I'm still searching for that. That's some very
useful advice, and I will have a think of that. Hopefully
I'll get some nice suggestions, but also I'll try and
think through what's the best way for me to get that
message out.
Kirill: Fantastic. Well, we'll leave it at that. Stratos, thank
you so much for coming on the show and sharing your
story with our listeners.
Stratos: Thank you so much.
Kirill: All right.
Stratos: Thank you so much, and, yeah, we'll speak to you
soon. Thank you so much for the invite again.
Kirill: So there you have it, everybody. I hope you enjoyed
this episode as much as I did. For me, the most
exciting and inspiring part was the dedication that
Stratos has. It takes a lot of courage and a lot of
commitment to buy a ticket and fly all the way from
the UK to Los Angeles just for three days, to attend the
event that is going to change your career. But as you
can see, Stratos didn't make a mistake. Stratos
actually made the right choice, and that helped him
follow in the path of the career of Rico Meinl. How
exciting is that? I wish the same to you.
Kirill: To wrap up this episode, as usual, you can get the
show notes for our conversation at
superdatascience.com/351. There you can get the
transcript for this episode, any materials that we
mention, and of course the URL to Stratos's LinkedIn
and places to connect with him. Highly recommend
connecting and staying in touch. He will surely be
happy to answer any questions you might have about
interviews, about creating a data science ecosystem,
about courses, about conferences, about podcasts. Get
in touch. Stratos sounds like an amazing guy who is
going to be able to help you, whatever your questions
are.
Kirill: On that note, thank you so much for being here today.
I look forward to seeing you back here next time, in
our next amazing episode. Until then, happy
analyzing!