Real Talk with Google Data Scientist (with a PhD in Physics)

[@KenJee_ds]

I think that the Mercedes problem is very interesting! This shows that data science isn't just about running huge volumes of data through an algorithm, there can still be computing power constraints in the world today. I have some videos about the different types of data science projects out there on my channel that would complement this one nicely!

[@healthfreak5438]

This is one of the best interviews among other videos on your channel, both interviewer and interviewee knew what they are talking about.

[@SunnyHyde]

Bro she was reading off a piece of paper 😂

[@atrimandal4324]

Hey, Vsauce! Michael here

[@designerspen7420]

LMAOOOOOOOO! bruh

[@777kidik]

lol

[@kashyapshirodkar7947]

Or is it?

[@MeghangNagavekar]

I actually clicked on this video thinking that's Michael 😂

[@andrewtran6669]

his name is actually michael 😂

[@dylanlivingston4925]

It is encouraging to hear of Michael's involvement at Mercedes-Benz. Having worked with these systems as an automotive technician, it fuels my desire to transition to data science knowing that my twelve years of automotive experience will not be a setback, but a great advantage in the industry. Thanks for sharing!

[@yujiaoli947]

5:55 How to develop intuition for data: modeling... Disagree with this. Intuition could start from basic data description (mean, var, distribution) , simple visualization (correlation, trend, heat map, etc.), expert discussion (know business knowledge and background). Then, one can use some tests to learn the features' importance. After having such first impression of data, one can have intuition. Modeling should be after these steps.

[@commentmachine1457]

Yujiao Li i also got quite some disagreement when he thought "data learning" can replace exploratory work, wangling process thus feature engineering. Data science is a lot more about prediction, eventually insights are the derived value, which comes from beating the data

[@melaniaadelia9387]

so if i want to be data analyst i must had good skill at statistic?

[@fredwu5965]

Now DS applicants need to compete with guys like PhD in particle physics, wow!

[@apoozeo1715]

Common theme in my observation over the past year. Look at the data science team of Wayfair for example. Neuroscience, biochemistry PhD's are now becoming Data scientists. If you think about it PhD guys have at least 3-4 stats heavy courses. But the ones who get calls have published 2-3 papers which have some large scale data analysis. The best way to get noticed is to do say 5 kaggle projects. Cover all areas like binary classification, imbalanced datasets, regression, CNN and lstm etc. Also, important to do some basic SQL like he said.

[@DistortedV12]

I@@apoozeo1715 If you want a really good job especially in today's climate, I think the answer is in your original answer, publish in top ML journals like at Neurips not just in Kaggle data science tournaments. The competition is changing.

[@patrickchevnenko5789]

actually machine learning is from academics top phd like that in the beginning, it just Andrew Ng that democratize and over-simplified the Machine Learning courses

[@DanielBoa7]

this has always been the case

[@elvis_mello]

At least in my country (I live in Brazil and I'm graduating in physics) being a physicist that really researches is hard, especially now that the government ended most of the financial incentives for it. But politcs apart, the graduation is giving a lot of advanced math knowledge together with the concepts needed to solve problems. Generally is easier to have a more enjoyable life if I just follow another profession since almost any would benefit from what I'm learning.

[@glowish1993]

I could listen to him for hours and learn so much

[@YiannisPi]

Great video and thanks for sharing. Just to add on the SQL Vs Python for data science, my view is that Data Scientist need to know SQL just because all the data-sets in all organisations are stored in an SQL-like database so you need to know how to access, clean, manipulate, join and retrieve the data. Python is not a database. Then, you need Python to model the data & run algorithms; which is what SQL cannot do (not that well anyway). Is that simple :)

[@OFresh]

agree 100%!

[@ManPursueExcellence]

Thanks for breaking things down. I get confused with all of this.

[@pranavdange5695]

Amazing. This is really very helpful. Working in the Data Science field, you really get to understand why he is stressing on PRACTICE. I also like the ease at which he is talking

[@TheKinsey06]

Surprisingly I understand everything he was saying. I must be on the right track.

[@treqqqq2850]

I thought he was the guy from ex machina for a moment!

[@konventionmedia2352]

So did I 😂 they look so much alike

[@Santoshsusarla]

true !!!

[@RunOs3]

Wow, this was a great interview. Thanks for posting and thanks to the guest.

[@Newkeralaexpress]

Wonderful explanations for data science and its opportunities...

[@DistortedV12]

"What attracted you to the field?" His answer:"enjoyed the techniques and tools" Real answer: MONEY&CLOUT

[@wangxu2155]

MONEY MONEY MONEY

[@l.1244]

Do they earn that much though?

[@qaisarnawaz7236]

You kidding me?? He wouldn't have gone for the PhD if he was THAT attracted to MONEY

[@qaisarnawaz7236]

@Intelligence Injection That is exactly what I am saying... If he wanted money, he would have gone to data science straight from undergrad not after doing a freaking PhD in Physics... It is really REALLY hard to get though a PhD and money is never a motivation for a PhD student

[@LordTufi]

Didn't realize white folks were clout chasers, too.

[@ssharma6339]

Imho, SQL is important. if you get flat files from a retail client that have been pulled straight from a RDBMS, you DO need sql to first run some queries to narrow down your result-set by performing multi-table joins and then load the results into a pandas dataframe. SQL is an important skillset even from the point of view of data engineering and building data pipelines.

[@shamikbanerjee1454]

This interview is a freaking goldmine.

[@my_j.a.r.v.i.s.]

The way he answered all the question, it tells he is genius at his work.

[@NuEnque]

Genius? He works for someone other than himself. See what "The TechLead" or "Jarvis" have to say about Google. He failed his first interview and i'll bet those people didn't have a PhD. Hell, he can't even score the chick sitting next to him. He just has focus. Definitely not distracted with chasing women or getting wasted with friends or watching sports. #facts

[@bc9273]

@@NuEnque wtf does trying to sleep with the interviewer have to do with anything? That statement shows how juvenile you are. The phrase "score the chick" also means you're probably 30+ still acting like that.

[@NuEnque]

@@bc9273 Was that all you got out of my comment? Clearly I struck a nerve!

[@bc9273]

@@NuEnque your comment was ridiculous. You tried to insult him to make up for something you apparently lack. I would say that he clearly struck a nerve with you. As for you striking a nerve with me, I don't get bothered by grown men using insults made by children on an elementary playground. I probably shouldn't have made my original comment, though. People your age acting like that are never going to learn from anything or anyone and mature.

[@gabriell7640]

NuEnque you're truly obnoxious

[@mohamed_v1]

" if you not practicing you can't develop that intuition "

[@emiliod90]

Agreed, time spent in feature engineering and actually building an automated pipeline

[@tuanva6484]

Micheal said that using SQL as same as using Python in data science?

[@malta-q4966]

he referring to web scraping. You can web scrap using python in a framework like ScraPy. But you can also use SQL to if you have direct sources of the data

[@hyesky26]

It’s amazing to see a data scientist wearing a digital large number watch 😁

[@SergioGomez-qe3kn]

It looks like a smart watch to me. Samsung Gear maybe.

[@ahmedbadal3795]

14 * 5 = 19::: change my mind.. i still use calcultor when it comes to addition ...and this have phd in pysicis

[@netoalves9211]

He is a such clever guy, he has an inspiring career and I would like to follow in his footsteps.

[@MrAbhithepandey]

What I know and think about Data Science and Machine Learning is correct. This is what I learned after watching this video.

[@domtorque]

Oh I see, didn't realise Babish had another job.

[@peregudovoleg]

Great tips. Thanks for the talk!

[@khwlel]

I loved the part about SQL because I have always thought that its easily replaceable, I got rejected for a job even though I ticked all the categories and more because there was someone who uses SQL more than I do. According to them of course unless they were just being nice :D

[@aarav1698]

Is springboard India is equivalent to springboard usa ?

[@nadiaz4769]

What are the best certifications for becoming a data scientist? I'm reading and doing exercises in crash course in Python by Eric Mathes and love it!

[@ai.simplified..]

4:56 wrangling will replaced?5:13 11:25 well said

[@21Gannu]

how can i get that car data set he mentioned.

[@21Gannu]

my email is [e-mail]

[@CasternFernandes22592]

archive.ics.uci.edu/ml/datasets/auto+mpg

[@prerakhere]

@@CasternFernandes22592 👍

[@shoebumx]

www.google.com/search?rlz=1C1GCEA_enIN833IN833&ei=mbt_XMGWD4200PEP_9-OkAg&q=uci+auto+mpg+dataset&oq=auto+mpg+uci+dataset&gs_l=psy-ab.3.0.0i22i30.635.3623..4456...0.0..0.599.6531.5-12......0....1..gws-wiz.......0i71j0i20i263j0.s-LtFG9jeao Click on the first link

[@jollyhrothgar]

@@CasternFernandes22592 Yep this is the one.

[@caringtonmathalephahlamohl7487]

If it helps, he is a machine learning engineer, not scientist. It comes from his reasoning about models. An engineer does bottom up approach... a scientist does top-down.

[@prasadkadam4915]

And which approach turns out to be optimal?

[@maxim9280]

@@prasadkadam4915 I think both are important. Id say an engineer is better at various tools and a scientist has more intuition and theoretical knowledge. Thats my opinion

[@ThyRiki]

@@prasadkadam4915 From experience, a scientist tends to do quite a considerable amount of research, and as such, does more theoretical work before getting his hands in the dough. Scientists' workflow tends, thus, to be considerably long until there is any valuable execution, with a promise of a very fine-tuned model upon completion. Engineers tend to first find something that works, regardless of how perfected it is, and then improve it in a somehow iterative way. Thus, this can take several things in consideration. Do you have the money to invest in such a long term project? Are you working on R&D or are you executing on a small time frame? Do you need the best model in the market, or a model that will get the job done? What is optimal depends in the company's resources and objectives.

[@mr.k.Gaming90462]

I saw his ads in my music organizer and it was soooo long and annoying I have the same ads from Springboard

[@908572813]

really nice to hear you talk. tnx

[@the_number_one]

SQL is Sequel?

[@keysangyonthan]

Sequel

[@Prizzzy]

Yes

[@the_number_one]

Static(?) Query Language

[@the_number_one]

Lazy programmer cliche

[@superawesh]

DID I HEAR THAT CORRECT ...... DATA LEARNING :O :O :O :O :O

[@akhilreddy1770]

That's deep learning I guess

[@JoeWong81]

Thanks for the interview

[@melissapenkala3613]

Good info. Thank you!

[@Rujuta1959]

After Phd physics u became data scientist .. my question is - so the time and knowledge u invested in physics was useless ? OR your knowledge of physics and knowledge of data science made good package for job? Or u shifted to new field?

[@banjo-beard]

Nah, it wasn't useless at all. My graduate studies built the technical foundation that has served me throughout my industry. It's not necessary to go to graduate school for science to switch to DS - it's just the path I took after completing my graduate studies and deciding to head for industry. We're all collections of experiences and influences - it's hard to underline something I did, and then say "yo, go do this". The things that helped me were learning to program, learning how to work with large data sets, tell stories with data, math and statistics. I basically didn't know any "ML" when I started my journey - that was all picked up on the job.

[@H1TMANactual]

Did he really say use LR to do classification, or did I mishear that?

[@robinbiron1790]

Linear regression for regression pb and logistic regression for classification pb. So you can use both LR to solve most data science pbs.

[@ak47ava]

Is it possible to be a data scientist without phd?? is it possible with just a bachelors degree?

[@prasadjayanti]

yes it is but it helps to have more experience & solid background in theory!

[@applewatch4085]

most data scientist are not PhDs! I have never been in industry (have a PhD in math), but that's what I get from Andrew Ng's speeches. PhDs (research scientist/ applied scientist. whatever the title may be!) usually do the research/problem solving/ and hand it down. It does not not seem the data science wave will be dampened for awhile, but I want the hotness and career opportunities decrease so I see what people will be "passionate" about!!!

[@melaniaadelia9387]

@@applewatch4085 hi im undergraduate business student n really want to have more study abt data analyst, can u recommend me kind of books or courses that worth to learn????

[@applewatch4085]

@@melaniaadelia9387 "The Elements of Statistical Learning" is very well known book, whose authors are among the top two teachers/explainers I have ever seen. It is freely available on their website. You may find it hard to follow, but good thing about it is that at the end of each chapter it has bibliography. So, you would be on the right track. UW machine learning course on Coursera is good start. For deep learning, Andrew Ng just gives big picture, you can look at "CS231n" on youtube, this is better than Andrew. Also, what follows should have been mentioned first: depends on what you want to do. research? or not Recommender system?, autonomous driving, etc. Perhaps you should talk to faculties at Stanford or UW who are on the cutting edge of science and also work for industry, and some data scientists in industries like Google Amazon, etc. Lots of them may not respond, it (talking to people, especially in this field, according to my experience) is a low-output process that you have to do, the sooner the better. Also, you can start working on Kaggle projects, real practice is really really important, and will instill the knowledge you learn in you! You will see problems that are not in the books which are vanilla. Let the problem lead you what is needed to learn. I, myself, am stuck behind the industry doors, however, I hope it was helpful!

[@melaniaadelia9387]

@@applewatch4085 thankyou very much, hope ur career always going well. God bless u😊

[@gopaljamnal7577]

Data Scientist, working in luxembourg but one day my dream will come true for working google as data scientist. I have PhD in IoT system, M.Sc in software engineering and B.Com in business. I work on random forest model and data visualization.

[@melaniaadelia9387]

hi im undegraduate business student n really want to know more abg career of data analyst, can u recommend books or course that worth to learn??

[@prosimulate]

Overqualified, for the job.

[@mohinichauhan4635]

That's not so true. One never received any word from Amazon, Facebook etc recruiter even after working for Google

[@mohinichauhan4635]

@Jerome Dillon Yeah, ur ryt ! human.I will definitely follow up on your words just to see if leaving google is worth for me :)

[@arrabalimaz622]

I love interviewer's fake smile continuously she was onn....

[@acidnynex]

Good job, bud! You're famous now!

[@iLeven713]

Highlander reference. Love it.

[@shekharyadav3462]

You guys are doing great job

[@christianherrera4729]

Does anyone have any good book recommendations that have helped you broadly in your career as a data scientist? (Not necessarily technical books)

[@Urbrosfriends2ndcuz]

The Master Algorithm is a good book for breadth if you are just starting or early in your career.

[@melaniaadelia9387]

@@Urbrosfriends2ndcuz is it pedro domingos books????

[@Urbrosfriends2ndcuz]

Yeah, that's the one :)

[@aination7302]

What is recall? Anyone please

[@banjo-beard]

Classify a data set with a rule or algorithm. Then, count the number of true positives. Count the number of false negatives. Divide the number of true positives by the sum of true positives and false negatives. That's recall! More intuitively, recall gives tells you what fraction of positives in a dataset your model can correctly identify.

[@flaskwater44]

This is the data set he references archive.ics.uci.edu/ml/datasets/Auto+MPG

[@ccladiesman]

What is recall? And how is that different from precision and accuracy?

[@Davesean3]

en.wikipedia.org/wiki/Precision_and_recall

[@anthnysalazar2887]

Real Data Scientist, Mathematician, statistician, physicist, engineer, lawyer, accountant, administrator or whatever the career you have through university title ARE nothing against real experience(not gotten throughout a company or university, there just exist parasites stealing money from the Corp and average people loosing time at some college )..... Make money, shine and live by yourself, if you can!!!.... most notable developments was not lead by if not(unknown ones later it is know but much much time late)

[@rodrigomeireles5966]

did you have a stroke? are you ok?

[@sovannayak2335]

Do I need a science background to learn data science.?

[@RT-md3zs]

Thank you so much for being here today.. You can leave for the day..🤣🤣

[@fahadreda3060]

Great Video , Thanks

[@xinyuyou7512]

Great video!

[@Q_QQ_Q]

*Should have been the process like what level of projects to be there to have job*

[@olemew]

so always use logistic regression, gotcha!

[@ProfessionalTycoons]

very good video thanks.

[@multigladiator384]

does he mean RAM when he says it does not fit into memory

[@avatar098]

Yep! In programming, when we say memory, we generally mean RAM.

[@sticklebrix1756]

It makes me laugh that they even bother mention the "transition" from Physics to Data Science, as if Physics does not already have a ton of maths it in and clearly shows someone already has the aptitude for it. Someone going from a PhD in art history to data science would be of note, but this...

[@MegaRockandroll09]

Actually this is something that is going around my head for a while,clearly you need to be good at maths(basic + advanced) and statistics to be a data scientist.I have seen courses that don't have maths and they just teach tools and stuffs,sure tools help, but sometimes you need to develop your own sets of formulas to solve a certain problems with respect to data science.

[@melaniaadelia9387]

@@MegaRockandroll09 hi im undergraduate business student n really want to have more study abt data analyst, can u recommend me kind of books or courses that worth to learn????

[@MegaRockandroll09]

@@melaniaadelia9387 , you need to know basic statistics and some some basic mathematics,for books you get books on platform(like R or python) which shows you how you can apply those maths functions to get the desired output.

[@468_raghavdodla3]

@@MegaRockandroll09 Bro, I too am a theoretical physics enthusiast, pursuing a dual degree from IIT Madras in Data Science and programming. What about you??

[@munsif2320]

I’m next data scientists of Google 🥸

[@maz4998]

Was watching on 2x speed, misheard his last line as 'thanks for letting me out' 😂

[@CardioCareWellness]

6:15 auto mpg dataset

[@liamboyd4676]

maybe find a host that can have a productive, insightful conversation with the guest? Some of the answers worth following up and drill down a bit.

[@arthurferdinand94]

Mantap sebagai.data science

[@husammuhanad2548]

Great info

[@rajath1964]

What is his full name ?

[@buildwithcode9743]

Michael Toblerone

[@username42]

use google :) i easily find him in 10secs

[@amandamate9117]

Michael Beaumier

[@username42]

@@amandamate9117 first you said his name is bald testestorne junior and then changed to beaumier ? :D

[@BlairBryngelson]

BRENT KASKEL

[@aryamahima3]

I'm a PhD in Theoritical optical physics. I am thinking about switching my career to data scientist.

[@mihir469]

Wow!

[@Stephen-hu8ec]

Did you switch?

[@f_ckaroundnfindout3915]

I feel you ,The job prospect in academia is really bad and it's hard to get a tenure nowaday. :(

[@nitishjoshi9456]

Please share his LinkedIn profile or full name

[@username42]

use google :D it is ez it took me 15secs to find him xD

[@nitishjoshi9456]

@@username42 thanks.

[@username42]

@@nitishjoshi9456 no problem

[@amandamate9117]

Michael Beaumier

[@apoozeo1715]

@@amandamate9117 this ☝

[@CatzNGaz]

I'd bet data learning doesn't sound so arduous for the entire DS&ML communities if a data sets variables exclusively have dependencies on one another, that can be learnedly identified and constructed into functions. However, there are often variables that do not exist in a data set, for which multiple features depend on. It'd be vacuous to contrive such free variables in a purely statistical sense, but I suppose PDE's lend some workarounds for evaluating possible behaviors of a free variable via setting realistic initial conditions.

[@vasanthkumar9488]

Tq

[@RJG253]

Awesome.

[@kiralight4212]

Nice!

[@zirokl]

useful.

[@marcpadilla1094]

Son of a gun is loaded.Brains and Voocoo cash. These techies are royalty nowadays.

[@NotFound-hy7qb]

Why he is focusing on just building models ? I mean shouldnt you try to understand your data first as model building is just like writing 2 lines and getting through it doesn't require much skill

[@sreevisakhjain]

Great

[@DistortedV12]

Not to offend this guy too much, but I wonder if he will pass the data science interviews of today. It seems that nowadays they are riddled questions to hire a non-existent ex-facebook core developer who took some time off to get a Ph.D in statistics. Exaggerating, but yeah...

[@tommyjado127]

thank God you put "PhD in physics" in the title omg how else we would respect this guy ? ~sarcasm~

[@yamirdreizehn283]

I guess you can only thing about yourself. As a physic student it is quite useful to know your future job paths.

[@bc9273]

Because it's an accomplishment he earned? He left a great field for data science, that's kind of important to know. Your jealousy is showing.

[@spinLOL533]

good video

[@pelecranileboi9011]

You look like michael stevens

[@umairalvi7382]

Story of students of different fields nowadays Arts->computer science Physics->computer science Electrical,mechanical,electronics....etc->computer science Psychology->computer science Biology->computer science . . . .computer science->what the heck im gonna do .

[@lizantoinetteanguren3125]

I feel the last one

[@AnhTuPhucDerrickHoangCanada]

I dropped DS as I'm more AdWords than AI.

[@sunsidhe]

If your first thought wasn't "Hey, that's the guy from Ex Machina", then what's wrong with you?

[@lekhapratap1652]

The sexy Doppler Effect.

[@jony7779]

"SQL vs Python: WHICH ONE IS THE RIGHT ONE ALWAYS WITHOUT EXCEPTION IN EARLY MARCH 2019 ???!?!"

[@learnthenew3119]

Go with Python it will be undergoing different Job fields like Big Data, Data Science, Web Technologies, and AI.

[@tigerrx7]

If you’re asking to choose between the two you haven’t come across the right project that’ll force you to use both heavily, SQL --> Python for post process.

[@MrPauricg]

They are two different technologies completely if you are doing data science at all you will need python. for building models , pre-processing, data wrangling all of the above. SQL will just be used if you have your data not stored in memory, I work as a data scientist and use python and R every day. 80% python. I learned some SQL while studying but have quite literally never used it in work

[@MegaRockandroll09]

@@MrPauricg i have questions,Do you need to be good at maths to be a data scientist??

[@melaniaadelia9387]

@@MrPauricg so what the things that always important as data analyst or business intelegent, im undergraduate mba n really want to be data analyst

[@Death_User666]

Sucks how you need a diploma

[@TheYaBoyKevin]

Thought about being a data scientist but forget getting a PHD. 😒

[@jeremywright9511]

Honestly you do not at all need a PHD to be a data scientist, though it is impressive. Some people get into the industry with much less; this man's qualifications are beyond the job's demands. Consider also that this is google rather than a standard company; they are a lot more selective and are not the standard for most.

[@smellypunks]

Highlander

[@philoneill9865]

I wonder if Richard Feynman would go into advertising if he lived today?

[@biggtk]

What a sad thought :(

[@BlairBryngelson]

CX SSJ 😁

[@NuEnque]

I thought you didn't need a degree to work at Google? Yeah right

[@applewatch4085]

career was the only reason, the rest reasons:BS

[@santiagoromerobrufau]

Apple Watch gotta find some deep-sounding reason, though.

[@TanishqIsHere]

Wassup with the potato microphone?

[@LyleGlenn]

Yes! And why does the interviewer have a better one than the interviewee?

[@MrFally2009]

11 dislikes must be from google 😂

[@thulsy0216]

I’m from Google and I like it :) Very insightful and spot-on. (My personal opinion, not representing Google.)

[@TheAcolossus]

bro you got more money so u bailed physics dont lie

[@anthonyarce8772]

he did say it but in coded language "good career opportunity"

[@haileywarner5109]

Can you really blame anyone for leaving academia

[@richardmiller2049]

So is Google working to silence conservative voices or shift conversations and search results the way other big tech is doing?

[@MrEmrys24]

Babish?!