B1 中級 美國腔 322 分類 收藏
I get paid a hundred thousand base salary,
like I'm hired as a new grad.
Hey guys welcome to this episode of
Reality vs Expectations.
Where I get people from different careers
and I have them write about five cards of what they
thought their career was gonna be like
versus what it's actually like.
Today I got Joma.
Currently he is a Data Scientist at Facebook.
Previously he has worked at LinkedIn
Buzzfeed and Microsoft.
He graduated from the University of Waterloo
with the degree in Computer Science.
Lets get this video started.
Hey guys today I got Joma.
He works as a Data Scientist at Facebook.
But Joma can you tell me
the story about how you became a Data Scientist?
Yeah sure.
So I did a lot of internships when I was in college.
I did some Software Engineering internships
and I also did a Data Scientist internship at Facebook.
And then after that,
I worked fulltime at Buzzfeed as a Data Scientist.
And then finally I came to Facebook as a Data Scientist.
Can you walk me through a day in your life
as a Data Scientist?
Yeah sure so mostly what we do is
We come into work
and then usually we have a lot of meetings.
Because we have to talk about
what are like the next goals
and what metrics to track for our team.
So for example I work for videos
at Facebook.
And basically what I do is I query a lot of
queries getting some data
to try to make decisions for the PMs.
Give me an example of a specific type of data
that you're looking for?
Yeah sure like for example
You wanna see alright what countries are doing
like most well for our shows.
Cuz now we have a watch tab on Facebook
And we wanna know which countries are doing the best
and which countries should we invest in.
And that's one way to look at it.
And how did you become
qualified to get a Data Science job?
Yeah so It's actually
a lot of different background
You could come from a lot of different backgrounds
My background is in Computer Science so
with the Computer Science degree
I was a little bit more technical
and then when I did the internship.
Because the internship
they allow anyone to get it you don't need to be technical.
And that's where I learn how to do
Data stuff like for example basic stats
Did you come from like extremely qualified school?
is that why your looking for internship?
I went to University of Waterloo
which is a Canadian school and they do a lot of internships.
I wouldn't say its like the best school in the world
like its not ideally at all
But we do a lot of internships and maybe that's why
we get more preference over other schools.
So lets go to the Reality vs Expectations questions.
What's your first card?
So my first card is
Most people think you need a Ph.D
to be a Data Scientist.
And that's actually a myth.
Because I don't have a Ph.D.
I should just have a bachelor.
And then at Facebook I've met many people
that have the Neuro Science degree
or even someone that had a
Family and Sexuality degree.
And most people come from consulting backgrounds
So yeah so you deffinitely do not need a Ph.D.
I think the reason why peope are confuse by it.
It's because when you think of Data Science
you think of the machine learning of Data Scientist
Since they came from different backgrounds
how did they teach themselves
or how did they learn the skill well enough to get a job?
So theres two ways
Either you do an internship
or you just study basic stats.
because to be honest
its less about learning the fundamentals
or like being really good at the stats
to be good at this Data Scientist.
You usually have to have more empathy
to be good at Data Scientist.
Because you have to ask the right questions
and then answer them thoroughly.
Because technically its not that hard.
You only need to know some sequel queries
maybe a little bit of Python which everyone can learn.
What's Reality vs Expectation card number two?
Yeah so this is a little bit related to the previous one.
Most people think Data Scientist
works so solely on machine learning
and like artificial intelligence.
But that's not true
I just wanna talk about the three arc types
of Data Scientist.
There's one
is Data Science analytics
that's what I am.
And then I'll talk about that later.
And then there's Data Engineers.
And then there's also Data Science Core.
That's what people at facebook calls.
So Data Science Analytics this is like us.
We just like a Data
We do some sequel queries we process it we make graphs
and we communicate with to the Product Managers
and then Data Engineers.
Those are the one that retrieves the data
build the infrastructures
So we can actually look at the data.
And then Data Science Core those people are like the
hardcore Ph.D with like recommendation models in forecasting.
Awesome, what's card number three?
Yeah so card number three is
Data Scientist is just about
putting a bunch of data in a Blackbox model
and then I would just output an answer.
So that's very not true because
What's most important in Data Science
is about you know like I said
empathy and also understand what the real questions are.
I'll give you an example..
Why you can't just put in a Blackbox
Now Blackbox what happens is you give it input
and then you say what to optimize for.
Now imagine you have a video product
and you wanna focus on a specific country
in the emerging market.
And then to help boost your video product
and then what you wanna optimize is time spent.
It makes sense right because you wanna
make people watch more videos.
And then you put in a Blackbox and it says oh
Vietnam or Thailand is the best country.
But then that doesn't tell you the whole story because what if
the reason why they spent so much time
It's because their just spending time buffering
or loading the video.
So that's why you can't just put things in the Blackbox
Cuz you have to understand exactly what's happening
to the users on the other side.
Can you define Blackbox?
Yeah so Blackbox meaning like
models that people pre create for example
A simple linear regression
or like a random forest or even
like a deep learning model.
Sometimes you can't solve problems just by
encoding data in these models,
so that's why I mean by Blackbox.
It kinda relates to
like everyone thinks correlation equals causation
like a two things correlate they think it's causing
Obviously this Data Scientist you know better that
just cause two things are correlating doesn't mean
that you know the causation.
Exactly so one of the biggest mistakes is you know
correlation vs causation.
And you will always find a correlation
and then you would optimize on that certain thing
thinking that it would
benefit the other thing for example time spent
correlates with likes for example.
And then what you see later on
is that maybe if you increase time spent
It doesn't necessary mean
the more likes you'll get cuz maybe you'll just get
wasting time spent and stuff like that
time spent there are lower quality.
Exactly, what's number 4?
You need to know Hadoop
Mapreduce and Spark
if you wanna be a Data Scientist.
Cuz these are like the buzzwords that you hear the most.
And that's not true at all.
Cuz I've never written a Mapreduce job in my life
I have but not at my job.
And the reason for that is that usually
the reason why you think you need these is because your
applying to startups.
And startups they don't have enough resources to
hire the three arc types of Data Scientist.
So Mapreduce, Hadoop and all of these things
those are usually the Data Engineers that work on this
or Software Engineers.
Cuz technically you don't need to know much about
data or statistics
to create these pipelines.
So the Reality vs Expectations is that
you don't need to know these things
When you thought you did.
So for example.
Working at Facebook especially because it's so big
They have three separate jobs for that.
You know they have the Data Science Analytics
the Data Engineers and the Data Science Core.
So Data Engineers would do all that
and you wouldn't even need to think about it
You can just focus on you know impact
and thinking about how to
you know
how to make a product better with the Product Managers.
If someone wanna to get to Data Science today
What website should they go to to learn more.
I personally don't use any websites.
and I wouldn't recommend websites.
I think you should definitely just try to get an internship.
And to be honest,
this a little bit harder
But if you do have a technical background
like computer science it would be better.
And if you still can't do it.
Maybe try to get a consulting job
and then move into Data Science.
Okay, what's the last card?
The better you are at statistics
the better Data Scientist you'll be.
So what I mean by better at statistics
we usually think about complicated models
advance forecasting techniques and stuff like that.
I just wanna to tell you a little bit
about what happen to my internship.
We had
five interns.
One of them did some hardcore forecasting thing
That's like really complicated.
But the only thing it forecasted was for example
The number of active users
for that specific product and maybe it was very accurate
but what is that give us for product.
What kind of like product recommendations does it give us.
It doesn't really give us anything.
so in the end it's not about how good your stats is
or how technical you are
Value can you add to the company
And that's what matters the most.
Because if you do many complicated things
and like a lot of machine learning stuff
But in the end your not giving any value to the company.
even if it's so cool even if it's like
like really
advance stuff it doesn't matter.
cuz I can do the same thing
with a simple logistic regression
or like a linear regression.
as long as it has impact to the company
then that's what your valued at.
It's interesting you are saying that
It's almost like working as a
as a developer too.
It's related that you can develop this complicated code
But if it hasn't have any functionality
then there's no point
like it's kinda like the difference between like
doing something theoretical
versus doing something practical.
Theoretical can get so complicated but we can't use it
then what's the point.
Right exactly so I mean.
A lot more then often
I see people over Engineer softwares
and then yeah it's really good and it's marginalized
but nobody can touch it.
Because they just don't understand how to use it
and that's useless.
Alright Joma last question.
If you could go back to the beginning of your career
you're freshmen you just graduated from high school
would you do Data Science all over again?
That's a little bit of a hard question because
I do enjoy Data Science now.
But I think there are somethings that I would like to do more
I always wanted to be a Product Manager
rather than a Data Scientist.
Unfortunately like through
this schooling that I had done
I didn't develop the skills as a Product Manager
I develop the skill as a Data Science or as a Software Engineer.
So if I had to redo it
I probobly would have focus more on like
the business side of things.
What did you see in your professional career were
now you would prefer to be a
Product Manager than a Data Scientist.
Yeah so I saw that Product Managers
they focus more on execution.
And they also get
more of the credit when things go well.
in terms of different products
like in some sense a Data Scientist
is like the right hand man of a Product Manager
a Product Manager is like a mini CEO.
So I always love thinking about products
and thinking about you know
new and innovative way to think about
you know how to reach users
and how to make their lives better.
But usually it's the PMs that have the final say.
I get paid a hundred thousand base salary.
like I'm hired as a new grad.
I get paid the minimum
I get a hundred twenty thousand equity
for four years that means like thirty thousand equity per year.
And then for the first year you get
thirty thousand dollar bonus for the first year.
And then you'll also get some random relocation bonus.
That's worth like fifteen thousand or something like that.
I hope you enjoy that interview with Joma.
If Data Science sounds like a profession
you might wanna start learning about.
Consider taking this course I link in the description
on Skillshare.
I'm your instructor Frank Kane,
and I spent over nine years at Amazon.com
and IMDB.com,
developing and managing some of their most famous features,
like recommended for you and
I think it's a great introductory course.
And with Skillshare you get access to
over eighteen thousand courses
for fifteen dollars a month
what a great deal.
And with that being said I'll see you guys next week bye.


臉書資料分析師告訴你,這份工作跟你想像的不同! (Data Scientist: Reality vs Expectations ($100k+ Starting Salary 2018))

322 分類 收藏
kiki 發佈於 2018 年 5 月 16 日    Angus 翻譯    Evangeline 審核
  1. 1. 單字查詢


  2. 2. 單句重複播放


  3. 3. 使用快速鍵


  4. 4. 關閉語言字幕


  5. 5. 內嵌播放器


  6. 6. 展開播放器


  1. 英文聽力測驗


  1. 點擊展開筆記本讓你看的更舒服

  1. UrbanDictionary 俚語字典整合查詢。一般字典查詢不到你滿意的解譯,不妨使用「俚語字典」,或許會讓你有滿意的答案喔