1 of 67

Ace the Data Interview

CMU Data Science Club

By Nick Singh

2 of 67

About Nick Singh

At UVA:

Studied Systems Engineering & Computer Science

Data Science Intern @ Office of Naval Intelligence

Data Engineering Intern @ Google

Early Career:

Growth Engineer at Facebook

Multiple Hats @ Location Analytics Startup SafeGraph

For Fun:

DJ Lil’ Singh & Drake’s #1 Fan

2

3 of 67

4 of 67

5 of 67

Agenda

  • How Portfolio Projects Help With Behavioral Interviews
  • What Technical Interviews Cover
    • Prob + Stats Questions
    • SQL + Coding Questions
    • Open-Ended Business/Product-Sense Questions
  • Whether Technical Interviews Are Easy or Hard?

By the end of this talk you’ll learn…

June 30, 2020

Presentation Title to be adjusted on the 1st master page

6 of 67

Behavioral Interviews &

Portfolio Projects

7 of 67

7

7

8 of 67

8

8

9 of 67

10 of 67

But there was one slight problem…..

June 30, 2020

Presentation Title to be adjusted on the 1st master page

11 of 67

Users signed up… then churned out :(

Why did they churn out?

How can get users to return to the game?

June 30, 2020

Presentation Title to be adjusted on the 1st master page

12 of 67

The Solution:

Growth Engineering!

(a combo of Product Analytics, A/B Testing, and Software Engineering)

June 30, 2020

Presentation Title to be adjusted on the 1st master page

13 of 67

2,000 Monthly Active Users Later…

I Had Fallen In Love

June 30, 2020

Presentation Title to be adjusted on the 1st master page

14 of 67

Behavioral Interview Questions:

Why Facebook?

Why Growth Engineering at Facebook?

15 of 67

16 of 67

Behavioral Interview Questions:

Tell Me About a Time where you…

Analyzed data to find an insight

Ran an A/B test

Built something for consumers

Learned something new

Ran into a technical challenge

Did data analysis which impacted the product roadmap

17 of 67

Biden Dad Joke Maker

Your project doesn’t have to be fancy!

SQL + Tableau can go a long way!

18 of 67

19 of 67

19

20 of 67

21 of 67

22 of 67

23 of 67

24 of 67

24

25 of 67

Are Technical Interviews Easy or Hard?

25

26 of 67

Has anyone done a technical interview before?

27 of 67

How many technical interviews have you done before?

Click Present with Slido or install our Chrome extension to activate this poll while presenting.

28 of 67

Technical Topics Interviews Often Cover

  • Probability
  • Statistics
  • ML
  • SQL
  • Python Coding
  • Business & Product-Sense
  • Open Ended Case Studies
  • Take-Home CHallenges

28

29 of 67

Core Concepts for Prob & Stat Questions

  • Probability:
    • Combinatorics
    • Probability Distributions
  • Statistics
    • Random Variables
    • Expectation & Variance
    • Hypothesis Testing (A/B Testing)

29

30 of 67

Common Patterns for Prob, Stat, & ML Questions

  • ELI5 - Explain Like I’m Five
    • “Explain a p-value or confidence interval to a non-technical audience?”

30

30

31 of 67

“Explain Like I’m Five” Tips

31

31

32 of 67

Common Patterns for Prob, Stat, & ML Questions

  • ELI5 - Explain Like I’m Five
    • “Explain a p-value or confidence interval to a non-technical audience?”

  • Questions derived from your past projects & work
    • “You used Linear Regression — why?
    • How did you validate its assumptions?
    • How did you measure its performance?
    • Did it overfit?”

32

33 of 67

Common Patterns for Prob, Stat, & ML Questions

  • ELI5 - Explain Like I’m Five
    • “Explain a p-value or confidence interval to a non-technical audience?”

  • Questions derived from your past projects & work
    • “You used Linear Regression — why?
    • How did you validate its assumptions?
    • How did you measure its performance?
    • Did it overfit?”

  • “Straight Up” Stats/Prob Questions

33

34 of 67

Examples of ‘Obvious’ Prob & Stat Questions:

Say you have a sample size of N. The margin of error for the sample size is 3. How many more samples do you need to decrease the margin of error to 0.3?

Say you have a deck of 50 cards made up of cards in 5 different colors, with 10 cards of each color, numbered 1 through 10. What is the probability that two cards you pick at random do not have the same color and are also not the same number?

A coin was flipped 1,000 times, and 550 times it showed heads. Do you think the coin is biased? Why or why not?

35 of 67

550 heads out of 1,000 coin tosses.

Do you think the coin is biased?

Why or why not?

36 of 67

550 heads out of 1000 coin tosses. Is the coin biased?

Start presenting to display the poll results on this slide.

37 of 67

A coin was flipped 1,000 times, and 550 times it showed heads. Do you think the coin is biased? Why or why not?

37

38 of 67

Common Patterns for Prob, Stat, & ML Questions

  • ELI5 - Explain Like I’m Five
    • “Explain a p-value or confidence interval to a non-technical audience?”

  • Questions derived from your past projects & work
    • “You used Linear Regression — why?
    • How did you validate its assumptions?
    • How did you measure its performance?
    • Did it overfit?”

  • “Straight Up” Stats/Prob Questions

  • Stats/Prob In A Business Context

38

39 of 67

40 of 67

41 of 67

Survey Design:

We want to conduct a survey to see how people perceive our Disney+ UI versus the Netflix UI (is it easy to use, is it beautiful, which one do they prefer, etc.)?

42 of 67

Survey Design:

We want to conduct a survey to see how people perceive our Disney+ UI versus the Netflix UI (is it easy to use, is it beautiful, which one do they prefer, etc.)?

How many people should we survey?

43 of 67

How many people should we survey?

Click Present with Slido or install our Chrome extension to activate this poll while presenting.

44 of 67

Survey Design:

We want to conduct a survey to see how people perceive our Disney+ UI versus the Netflix UI (is it easy to use, is it beautiful, which one do they prefer, etc.)?

Why can’t we just survey a TON of people?

45 of 67

Survey Design:

How many people should we survey?

46 of 67

How well do you know SQL?

Click Present with Slido or install our Chrome extension to activate this poll while presenting.

47 of 67

48 of 67

https://datalemur.com/questions/sql-histogram-tweets

49 of 67

https://datalemur.com/questions/sql-histogram-tweets

50 of 67

50

51 of 67

51

52 of 67

52

52

53 of 67

The poll was deleted

Click Present with Slido or install our Chrome extension to activate this poll while presenting.

54 of 67

Open-Ended

Analytics Case Interviews

55 of 67

Open Ended Case Study

  • Often incorporates business & product sense, modelling, and domain knowledge
  • Example:
    • How would you build Uber’s surge pricing algorithm?
    • Watchtime is down on YouTube in India by 5%, how would you investigate the root-cause of the issue?
    • What dashboard would you build to measure the health of Facebook Marketplace?
  • Problem-Solving Strategy
    • Clarify, Clarify, Clarify
    • Align to Product + Business Goal
    • Mention Tradeoffs

55

55

56 of 67

Open Ended Case Study: Applied to Facebook Dating

What metrics would you use to define the success of Facebook Dating?

  • Clarify, Clarify, Clarify
    • Who is this product for? Brazil 18-50 year olds.
    • How does the product work? Hinge-style app…
  • Align to Product + Business Goal
    • How does this fit in with Facebook’s mission to “Give people the power to build community and bring the world closer together”?
    • How does it help Facebook make money?
  • Mention Tradeoffs

56

56

57 of 67

What metrics would you use to define the success of Facebook Dating?

Click Present with Slido or install our Chrome extension to activate this poll while presenting.

58 of 67

Uber Interview Question

What metrics would you track to make sure Uber’s surge pricing algorithm was working well?

58

58

59 of 67

How would you build Uber’s surge pricing algorithm?

  • Uber Quarterly Reports
  • Use Uber App: Both Rider + Driver App

59

59

60 of 67

“ETAs spiked to 8 minutes and the share of trip requests unfulfilled rose to 25 percent”

61 of 67

Are Technical Interviews Easy or Hard?

61

62 of 67

Technical Interviews Are Hard

  • Nuanced (churn is usually bad, but maybe okay for dating apps)
  • Company Specific (need to know about Uber or Facebook business model)
  • Time Pressure (SQL questions require speed)
  • Wide Range of Skills Tested (Coding, Stats, ML, etc.)
  • Require Non-Obvious Application of Skills (ie. 500 coin toss)

63 of 67

Technical Interviews Are Easy

  • Nuanced (but patterns can be learned with enough practice)
  • Company Specific (but business models, like marketplaces, are common + standard and you can learn main models via practice)
  • Time Pressure (but practice makes you much faster)
  • Wide Range of Skills Tested (but if you practice can learn main concepts for each topic)
  • Require Non-Obvious Application of Skills (but interview practice helps you apply your skills better)

64 of 67

Technical Interviews Are Easy

  • Nuanced (but patterns can be learned with enough practice)
  • Company Specific (but business models, like marketplaces, are common + standard and you can learn main models via practice)
  • Time Pressure (but practice makes you much faster)
  • Wide Range of Skills Tested (but if you practice can learn main concepts for each topic)
  • Require Non-Obvious Application of Skills (but interview practice helps you apply your skills better)

65 of 67

Technical Interviews

Are Hard

But with practice it becomes

Easier

65

66 of 67

67 of 67

Audience Q&A

Start presenting to display the audience questions on this slide.