You want to get into data science, right?

well, you're in luck because I work as a data scientist at

Untitled company wait I work as a data science at untitled company

Hey guys welcome hey guys welcome to another episode of

Hey guys welcome to another episode of Joma Tech.

Now you might be wondering why am I in my room?

Well, that's just how things are gonna be from here on out so quick little update

ummm....I had to remove my day in a life video

You know the one that got me most of my views. I had to take it down because I violated some policies so

That's why it's gone, but today. We're gonna be talking about something different

We're gonna be talking about how to get a data science job at a you know a large company like

Friendster or MySpace something like that first of all what is data science because I have a feeling a lot of people have you know?

misconceptions about data science so if you do I would suggest you to check this video I made with

Engineer truth so yeah check that out

And hopefully it will give you an understanding of what data science actually is or what my role is alright

So let's get to it to the five tips on how to plan a data science job number one tip

do an internship and have a technical degree how I got my first hands-on experience on data science was my

internship, and I did my internship in a large untitled company and

How I got that internship was was I just had a few software engineering background, which you can

See you can see my other internships on the previous videos, but yeah, I had a few internships

That's software engineering and my degree is in CS

I never touch the data science, but or so that's the thing so the easiest way is to get an internship

And then usually it will give you a full time offer right after so here are some examples of good degrees to have CS

software engineering economics statistics

math

Just engineering in general like system engineering

Maybe even environmental engineering just anything with the word engineering in it

It would you know demoness technical and they will give you an interview all right second tip?

Learn sequel or SQL. I'm not sure so when you're gonna start your work as a data scientist

No matter where it is you're always gonna have to write sequel because you got to get the data somehow no matter

How fancy your you know Python is and stuff like that in the end

Usually most companies set up their infrastructure so that you query data using sequel why because it's an easy language and even like

Analysts, or like business intelligence people they use sequel

So you know you'll probably have to use sequel, too

because it's just easier for them and then because they set up the infrastructure like that you have to use that too and

You also should learn sequel because all of the interviews. I got that was related to data science. I had a sequel question

Let me show you an example of question, so I found this take home exam

And I will change it a little bit so that the large company which is not the original

Untitled large company, but another large company that might or might not be related to transportation

Gave me this take home exam, so I'm gonna whip it out so for example what they would do is

They would give you a schema

and

They would give you a schema for example you see ID

Client ID driver ID city ID client rating driver rating and stuff like that and then the question would be something like

for each of the cities, San Francisco and

Los Angeles calculate 90 percentile difference between actual and predicted ETA

For all completed trip within the last 30 days, and then what you do is you have to write queries for that

Yeah, I'm not gonna go through the answer

But you should that would give you an idea of what kind of query you should run tip number three

Learn about success metrics and tracking metrics that is very very important

And it's kind of hard because it's more about product sense, so why is it important?

It's important because when you become a data scientist you will be the extrovert on your team on data

Right so imagine if your team is like at McDonald's, and you know your domain data scientists

then you're the one that will know where everything comes from like all the data sources and

Also, what to measure for success because you're gonna be the one that the PM's asks you like okay?

How do we know we're doing a good job?

How do we know that the introduction of these new hamburgers?

That they're actually successful and making mcdonald a better place for people you know sometimes. It's easy sometimes

It's a lot harder to define these success metrics. I'll give you an example taking from the same take home exam

so

Imagine if the YouTube creator app in case you don't know what the cue tube creator app

It's a it's an app for YouTube creators like myself and when you click on it

You could see like how many views are getting you know and how many

Watch time and like what video is performing a lot, what comments. Are yeah, it's basically a tool for creators like myself

So that's basically what the YouTube creator app is so imagine the YouTube creator app is being redesigned

You know you're adding CTR metrics are adding better interfaces

And they had like a I don't just better user interface like a whole facelift

How do you know that the redesign was successful and what are just success metrics in the tracking metrics alright?

Let me talk to you about the difference between success metrics and tracking metrics

Success metrics is usually the one that the whole team cares about you know they're always seeing growth usually at a start-up

It would be like how many active users you have for example not Viacom bonded a startup cannot find com or you can share

6.1

Second videos so imagine a Viacom I would care about how many active users like on a daily basis

How many unique people?

Actually open the app you know I would care about that would be my success metrics because that's the main thing I care about

For now and yeah, and then tracking metrics

I want to make sure you know that they're actually you know one real people or two having a good time on my app

So what are the tracking metrics? So that would be for example this think about it um?

I want to know if they're actually watching a lot of you know videos so tracking metrics can be watch time per active user

then or maybe just watch time in general and

Then another thing can be like how many videos are they watching you know

I want to make sure that the inter not just opening the app and doing nothing and then also

Let's see another tracking metrics would be you know how many

how many like force quit do they have why do I care about that because I want to make sure I want to track the

Negative stuff to like or my like or my apps being buggy like if they are and they keep force quitting

Then maybe the reason why they spend so much time

On the app is because it's really buggy right so I have to track these things to make sure that it's not buggy

Basically you want your metrics to be non treatable and represent success on your company, and then the next question

After asking like how would you you know define your success metrics? It would be how do you test it okay?

so there's one thing you need to know you need to know about a/b tests you need to know about control groups and

experimental groups and

Like how many people do you need in them enough so that enough so that statistically significant that you could see the differences

but also to reduce risk of

Having a shitty product that you launched right because imagine if you're the the launch of the new product or like a feature that you

Built on not fine calm is incredibly shitty. You don't want everyone

You don't want the whole 50 percent of your users to be exposed to that experiment right so you have to think about ways to

To make sure you don't risk that but also to get enough information that statistically significant all right tip number four

Learn ipython our tableau or excel or whatever to manipulate visualize and interpret data

So why do you need to know all this to be honest in my personal opinion a lot of people like to use ipython are

because they actually look really you know a

Technical and coding intensive, but you know honestly. I see a lot of good data

Science only use Excel right they just grab the data using sequel

They manipulated using sequel, and then you should put in Excel to visualize things because the impact is not how?

Technical you are but how well you communicate those findings, and how well translate into product recommendations now

I'll talk about communication later

Why do you need to learn all of these because you know you have to manipulate data

Time to sequel can't cut it and you need to visualize the data

Because you need to communicate them later, so so imagine you have a lot of data and your company

Let's see let's go back to McDonald's, so what do you do you do an exploratory analysis?

So imagine your McDonald's and you have a bunch of data?

About your sandwiches or your hamburgers and stuff like that so what the first thing you can do you could you know use it?

To download all the data

manipulate them group them by like you know hamburgers or

Day of the week and stuff like that and you could discover a lot of things using that and how would you do it you?

Would use either tableau Excel R or?

Python to manipulate things right because you want to answer questions like all right which day of the week

We have the most sales and now which hamburgers

Have you know which hamburgers are the most popular?

But then you could also say like which hamburgers actually have the highest profit margin plus

You know sales like so which one gives you more revenue

You know and stuff like that and then which ones are actually losing money because apparently when I work there

I scream every time

They sell ice creams are actually losing money because employees didn't know how to you know stop putting ice cream

So they filled it up too much

so they try to teach us how to

Put ice cream, but only put around it and leave a big void in the middle so that they save more ice cream

I never did that because I cared about my customers, so I want them to have the best ice cream ever even if

McDonald's losing money just saying so this is why you have to learn stuff like ipython are and stuff like that

Yeah, honestly it doesn't matter which one just as long as you're fast and you're correct at you know creating these

Visualizations, so yeah so Indiana use any of these technologies to show your results to interpret them and yeah

So that's why you need them alright number five

You have to learn to communicate really really well because the whole point of a data scientist is to communicate their findings

Right because if you don't communicate your findings. No one will know and if no one will know it's not gonna

It's not gonna flourish into anything. It's not gonna Forge into any products learn how to do public speaking

You know because you're gonna be doing a lot of presentations

You're gonna be talking to a lot of people so a lot of cross-functional partners a lot of people like PM's

Engineers and all that you're gonna be talking them all the time and you can be presenting to them learn how to write better

Because you can't be presidenting to everyone in the company

But you know whatever you found is valuable for maybe for another team, so how do you do that you write about it?

You know you write power points you are you write an article within your company you write a post or an email?

That's pretty important too because like writing is a very good medium to reach as many people as possible

especially at work learn how to make interesting powerpoints because

When you're gonna be presenting to people or when you're gonna be presenting you know

Your article you want to have graphs and stuff like that

And then you won't you don't want it to make it boring you want it to make it obvious what they're looking at I mean

That's all I see a lot of data scientists

They have like grass with like a shit ton of like fancy crap

But I can't interpret it like I look at it, and there's shit ton of bars and like colors and stuff like that

But what am I supposed to look at dude? I don't have attention span you know just like a video creator you need to

You need to get their attention within 2 5 second. That's the same thing with work if people don't want to read your shit

You're not gonna have any value added to them because they didn't read your post so make sure that when you present a graph

That it tells exactly what you're supposed to tell so enough information and it's obvious what you're trying to say and improves your thesis

So we have the same actually I just say that I don't think it's actually saying

But here's the thing if you made an interesting analysis, but no one is there to read your analysis. Did you really make an analysis?

Food for thought

But yeah, and then bonus last one learn predictive modeling, so this is not super required

because well

It's funny because most people these

Optimized on this learn predictive modeling whenever do you think data signs because they think data signs are really related to like

forecasting or like finance to finance

It would be pretty important because the whole goal of your of your job is to make more money

And how do you make more money you predict the future so that you can capitalize on?

But for us as like a product data scientists

Predictive modeling isn't that important, but it's good to know and it makes you a little bit cooler, so yeah

So it might be useful when you're like forecasting goals and stuff like that or when you want to build

recommendation systems on your product, but it's not super important, but it adds credibility and

also, you know

So that you can tell your friends. I you do modeling whoo all right, so yeah, so that's about it