Uber Data Scientist Interview

Uber's data scientist interview goes deeper on marketplace experimentation than peer company interviews, with rounds built around A/B test design in Uber's two-sided marketplace. Interviewers engage collaboratively throughout the loop, questioning assumptions and trade-offs in real time. A separate Bar Raiser round shapes your final leveling and carries more weight than any other behavioral evaluation in the process.

This guide breaks down each stage of the loop, what Uber interviewers look for, and how to prepare with real example questions, actionable tips, and resources.

Uber data scientist interview process

Uber's data scientist interview loop runs four stages over 3-6 weeks, with the onsite structured around SQL, statistics, product sense, and behavioral evaluation. The process varies by team; some teams include a take-home assignment between the technical screen and onsite, while others skip it entirely.

Here's an example of what the interview process can look like:

• Recruiter screen: A 30-minute call covering background, level alignment, and team fit
• Hiring manager screen: A 30-45 minute conversation on past projects and product intuition
• Technical phone screen: A 45-60 minute live coding session on SQL, Python, and statistics
• Virtual onsite: Four rounds at 45-60 minutes each, covering SQL, experimentation, product sense, and behavioral evaluation, plus a Bar Raiser

After the onsite, interviewers submit independent written feedback and the Bar Raiser moderates a debrief. Decisions typically arrive within 3-5 business days.

Recruiter screen

The Uber data scientist recruiter screen is a 30-minute call focused on resume fit, level calibration, and salary expectations. Expect recruiters to walk through your background and ask about your interest in Uber specifically.

Use this call to learn your exact loop composition. Uber recruiters are transparent about which rounds you'll face and what each evaluates, so ask directly.

Interviewers look for:

• Level alignment: Whether your years of experience and scope of past work match the target level
• Uber-specific motivation: A clear articulation of why you want to work on Uber's marketplace problems specifically
• Communication clarity: Your ability to walk through your background with quantified outcomes
• Logistical fit: Confirmation of location, hybrid schedule (Tuesday, Wednesday, Thursday in office), and timeline expectations

Hiring manager screen

Uber's hiring manager screen for the data scientist interview tests whether you can connect past data work to business outcomes. The conversation runs 30-45 minutes and typically includes at least one short product-sense question alongside the project discussion.

Interviewers look for:

• Business impact framing: Whether you describe past projects in terms of outcomes, metrics moved, and decisions influenced
• Product curiosity: Whether you ask 'why does this matter to the user?' before diving into methodology
• Team fit: Whether your working style and interests align with the specific team's problem space
• Scope calibration: Evidence that your past project scope matches the target level and what the team is working on

Technical phone screen

The Uber data scientist technical phone screen is a 45-60 minute live coding session conducted on CoderPad or CodeSignal, split between SQL, Python, and a brief statistics discussion. The format is consistent across teams: roughly 30 minutes on two SQL questions and one Python problem, followed by ~15 minutes of A/B testing conversation.

SQL questions lean heavily on window functions (ROW_NUMBER, LEAD, LAG, PARTITION BY) and require clean handling of edge cases like duplicate records, timezone conversions, and null values. The Python component is lighter, typically a Pandas data manipulation task or a straightforward algorithmic challenge.

Interviewers look for:

• SQL fluency under pressure: Your ability to write correct, efficient queries against multi-table schemas in real time
• Edge-case awareness: Whether you identify ambiguities in the problem statement and clarify assumptions before writing code
• Statistical reasoning: A concise explanation of when and why you'd use specific test designs, including sample size considerations
• Communication during coding: Your habit of narrating assumptions and trade-offs as you write

Sample questions

Here are some real interview questions reported by candidates:

• Write SQL to get the third transaction of every user. Output user_id, spend, and transaction date.
• How many customers who signed up in January 2020 had a combined sending and receiving volume greater than $100 in their first 30 days?
• Rank total distance traveled by each user using users and rides tables.
• Write a query to compute weekly chargeback rate by rider cohort, handling late-arriving chargebacks and duplicate disputes.
• Given a dataset of completed trips, calculate mean fare by city. How do you handle trips with null fare amounts?

SQL and coding round

The Uber data scientist SQL onsite round builds on the technical phone screen with more complex, multi-table queries tied to Uber's business domain. The round runs 60 minutes and covers scenarios involving ride data, payment transactions, driver activity, and marketplace metrics.

Some loops include a Python component in this round, typically a Pandas manipulation task or a short algorithmic challenge, though SQL is the primary focus.

The round typically pairs you with a senior data scientist. Questions about cohort retention, fraud detection, or metric investigation often require multiple CTEs, self-joins, or window function chains.

Interviewers look for:

• Query architecture: Your ability to break complex analytical questions into clean, readable CTEs rather than deeply nested subqueries
• Domain translation: Whether you can convert a business question ("Which drivers are likely churning?") into a precise SQL specification before writing code
• Performance awareness: Recognition of when a query approach will be expensive at Uber's scale and what alternatives exist
• Correctness under complexity: Accurate results on problems involving self-joins, multiple aggregation levels, and conditional filtering logic

Sample questions

Here are some real interview questions reported by candidates:

• Write SQL to detect potentially fraudulent driver behavior based on trip patterns.
• Find the most-used vehicle type by Uber customers in the past year.
• Get the distribution of total push notifications sent before a user converts.
• Return every employee who has ever worked for the company; for each, calculate the greatest number of employees during their tenure and the first date that number was reached.
• Compute rider retention by signup cohort. What percentage of riders who signed up in each month take a second trip within 7 days?

Statistics and experimentation round

The Uber data scientist statistics and experimentation round is the most distinctive part of the loop, and the round where Uber's interview diverges most from peers like Meta or Google. Uber's experimentation platform processes over 10 million experiment evaluations per second and runs thousands of parameters and experiments at any given time, and data scientists own the design, analysis, and interpretation of those experiments.

Expect questions that go well beyond textbook A/B testing. Uber's marketplace means that changes to one side (riders) affect the other (drivers), and standard user-level randomization often fails for interventions like surge pricing or dispatch matching. Interviewers test whether you can identify where an experiment design breaks and propose alternatives.

Interviewers look for:

• CUPED fluency: Whether you can explain variance reduction using pre-experiment covariates and articulate when it helps most
• Switchback experiment design: Your ability to propose time-based or geo-based randomization for marketplace-level interventions where user-level A/B tests fail
• Interference reasoning: How you identify and address spillover, network effects, and SUTVA violations specific to two-sided markets
• Practical judgment: Your instinct for choosing between parametric tests, bootstrapping, and Bayesian methods based on sample size and distributional constraints
• Multiple testing awareness: Whether you flag correction needs when multiple hypotheses are tested simultaneously and a single result appears significant

Sample questions

Here are some real interview questions reported by candidates:

• Your manager tested 20 pricing strategies and one came back significant. What would you do?
• When would you use a switchback experiment instead of a user-level A/B test? Walk through the design.
• Explain CUPED. Why does it work, and when does it help most?
• How do network effects break standard A/B testing on Uber's platform? How would you redesign the experiment?
• Sample size is too small for a standard A/B test in a new market. What alternatives do you propose?
• Define the Central Limit Theorem and explain its relevance for Uber's experimentation platform.

Product sense and case study round

Uber's data scientist product sense round pairs you with a product manager for a 60-minute case discussion tied to real marketplace decisions. Every case maps to an actual Uber business problem: pricing, retention, marketplace health, or new product launches.

Interviewers look for:

• Metric definition clarity: Your ability to define a primary success metric, explain why it's the right one, and name guardrails that prevent gaming
• Two-sided marketplace reasoning: Whether you account for supply-side and demand-side consequences for every proposed intervention
• Experiment design integration: Whether you connect a product hypothesis to a testable experiment with clear success criteria
• Second-order thinking: Whether you anticipate downstream effects (cannibalization, adverse selection, behavioral shifts) before being prompted

Sample questions

Here are some real interview questions reported by candidates:

• Evaluate $1, $3, and $5 cancellation fees. Which optimizes the marketplace? Define your primary metric and guardrails.
• Driver active hours dropped 10% in Chicago. Walk through your investigation framework.
• How would you measure the success of Uber One?
• Surge complaints increased 30% this quarter. Diagnose potential causes and propose metrics to monitor.
• How would you anticipate the impact of a large event (concert, stadium game) on demand in a specific market?
• Design metrics for a redesigned Driver app and propose an A/B test framework to evaluate it.

Behavioral and Bar Raiser round

The Uber data scientist behavioral round includes a Bar Raiser interview, a leveling round conducted by a senior leader from outside the hiring organization. The Bar Raiser often opens with a high-level strategic question before pivoting to behavioral questions. Prepare for both in the same round.

Unlike Amazon's Bar Raiser, which carries a direct veto on hiring decisions, Uber's Bar Raiser primarily influences leveling. A weak performance at the senior level typically results in a down-level offer rather than an outright rejection; for mid- and low-level roles, a strong no from the Bar Raiser is more likely to end the process.

Interviewers look for:

• Ownership signal: Whether your stories demonstrate independent decision-making, accountability for outcomes, and bias toward action
• Stakeholder navigation: Your ability to describe influencing decisions through data without formal authority, especially with senior cross-functional partners
• Failure resilience: A concrete example of a failed analysis or experiment where you articulate what went wrong, what you learned, and what you changed
• Values alignment: A clear connection between your working style and Uber's cultural values

Sample questions

Here are some real interview questions reported by candidates:

• Tell me about a failed analysis or experiment. What did you learn?
• Describe a time you influenced a decision through data without having direct authority.
• Tell me about a time you pushed back on something you believed was wrong.
• What would you do if a cross-functional colleague disagreed with your analytical approach?
• Tell me about a critical data quality issue you identified and how you resolved it.
• How have you worked with designers or product managers in past roles?

How to prepare for the Uber data scientist interview

1. Build SQL fluency on Uber-domain problems: Work through window function challenges (ROW_NUMBER, LEAD, LAG, PARTITION BY) using ride, payment, and marketplace schemas. Target solving medium-difficulty challenges in under 12 minutes and hard challenges in under 20.
2. Master marketplace experimentation design: Study CUPED for variance reduction and switchback experiments for marketplace-level interventions. Understand why user-level A/B tests fail for algorithmic changes like surge pricing and dispatch matching, and be ready to propose alternatives.
3. Practice two-sided product cases: For every case, name a single primary metric, two to three guardrails, and reason through the impact on both sides of the marketplace. Practice at least 10 marketplace cases before your onsite.
4. Read Uber's engineering blog: Posts on the experimentation platform, Michelangelo, and causal inference give you vocabulary (XP, Morpheus, Palette, switchback) that signals domain familiarity during interviews.
5. Prepare structured stories mapped to Uber's values: Each story should be under 90 seconds, lead with "I," and end with a quantified result. Have a strong "failed experiment" story ready; Uber's culture explicitly values learning from failure.
6. Confirm your loop composition with your recruiter: Uber recruiters are transparent about the process. Ask which rounds you'll face, who conducts each one, and whether your loop includes a take-home.
7. Run full mock interviews under realistic conditions: Practice live coding on CoderPad or CodeSignal without autocomplete, and simulate back-to-back rounds to build onsite stamina. Mock interviews with feedback help you calibrate pacing and communication in real time; for targeted prep with someone who knows the Uber loop, consider expert coaching.

About the Uber data scientist role

Uber's data science team owns experimentation and metric strategy across the company, embedded within product, engineering, operations, and policy teams. Data scientists at Uber own experiment design, causal inference, root-cause investigation, and metric definition; they do not own ML model development, which sits with Applied Scientists and ML Engineers.

Uber data scientists typically work on:

• Marketplace optimization: Pricing, surge, matching algorithms, supply-demand balancing, and incentive design across Rides, Eats, and Freight
• Experimentation platform support: Designing and analyzing A/B tests, switchback experiments, and geo-experiments across a platform that processes over 10 million evaluations per second
• Product analytics: Defining success metrics for new features, monitoring product health, and conducting root-cause investigations for metric movements
• Risk and trust: Fraud detection frameworks, payment risk scoring, and safety analytics
• Autonomous vehicles: Measurement and experimentation for autonomous dispatch alongside human supply, an emerging problem area following Uber's 2026 robotaxi partnerships with Nuro and Lucid

Uber's data science tech stack centers on SQL (Hive and Presto), Python (NumPy, Pandas, SciPy), Apache Spark for batch workloads, and Uber's internal experimentation platform. Familiarity with CUPED, switchback experiment design, and the scale of Uber's data infrastructure signals domain readiness.

Uber requires three days per week in office (Tuesday through Thursday) at hub locations including San Francisco, Seattle, and New York. New fully remote roles are approved sparingly.

Uber data scientist experience requirements

Uber senior data scientist roles (L5a) require 5+ years of industry experience in data science, applied statistics, or a quantitative field. A PhD can substitute for some experience, and strong L4 (Data Scientist II) candidates with 3+ years are considered for mid-level roles.

Proficiency in SQL and Python, experience with statistical experimentation and causal inference, and the ability to communicate findings to non-technical stakeholders are recommended. A background in marketplace, platform, or two-sided business models is preferred for Rides and Eats teams.

Additional resources

• Data science interview course
• Python coding questions
• SQL interview questions
• Statistics and experimentation questions
• Product sense questions
• Uber interview process
• Uber interview questions

FAQs about the Uber data scientist interview

How hard is the Uber data scientist interview?

The Uber data scientist interview is considered difficult because of its emphasis on marketplace-specific experimentation design, SQL fluency, and product sense. Four onsite rounds run in a single day, and the Bar Raiser round can directly affect your leveling.

How long does the Uber data scientist interview process take?

The Uber data scientist interview process typically takes 3-6 weeks from recruiter screen to offer, with most candidates finishing in about four weeks. Decisions after the onsite usually arrive within 3-5 business days. Timelines can extend if your loop includes a take-home assignment or if team-matching discussions follow the debrief.

How much does an Uber data scientist make?

Here are the reported compensation ranges by level for Uber data scientists, according to Levels.fyi:

• Data Scientist I (L3): ~$145K
• Data Scientist II (L4): ~$243K
• Senior Data Scientist I (L5a): ~$366K
• Senior Data Scientist II (L5b): ~$468K

Uber's compensation structure combines base salary, annual bonus, and RSUs on a four-year vesting schedule. Equity makes up the majority of compensation at senior levels.

Does Uber use a take-home assignment for data scientist candidates?

Some Uber teams include a take-home assignment between the technical phone screen and the onsite, but most data scientist interview loops skip it. When present, the take-home runs about a week, covers three sections (SQL analytics, experimentation design, and a predictive modeling task), and is graded by a data scientist outside the hiring team. Take-homes are more common on experimentation, forecasting, advanced analytics, and marketing analytics teams. Confirm with your recruiter whether your loop includes one.