Interview100x

Interview prep

How to prepare for an Uber data scientist interview

Uber's Sciences interview guide covers data science, applied science and scientist roles together, and lists eight competencies it may assess, while saying no candidate is assessed on all of them2. The loop it describes is long: live coding, a Jam Session, then four to six meet-the-team sessions2. The first job of your preparation is to find out which competencies apply to your role, then rehearse a full-length run so your answers stay sharp late in the loop.

Last reviewed November 12, 2026

How the interview process works

The Sciences guide lists a 30-minute recruiter call, 60 minutes of live coding, a 60-minute Jam Session, four to six 45-minute meet-the-team sessions, and a final decision, all virtual2.
  1. 1.Recruiter or sourcer call

    Virtual, about 30 minutes2. This is where to ask which of the guide's competencies your role is assessed on, which Uber itself suggests confirming2.

  2. 2.Live coding with a team member

    Virtual, about 60 minutes2. The guide's competency list names SQL, Python and data structures and algorithms under coding2, so ask which of these your session will use.

  3. 3.Jam Session

    Virtual, about 60 minutes2. Jam Session is Uber's own name for this stage, and the guide gives its name and length only, so ask your recruiter what to expect rather than relying on second-hand descriptions.

  4. 4.Meet the team

    Four to six virtual sessions of about 45 minutes each2. That can mean several hours of interviews, so plan for energy as well as content.

  5. 5.Decision

    A final decision after the loop2. Uber's general hiring page says the hiring team reviews performance against the role's criteria and competencies and against Uber's values1.

What this is based on

Uber's Sciences guide, as reviewed on 12 November 2026, gives stages, durations and a competency list, but no description of the Jam Session format2. Uber says steps can vary by role and location2.

Practice questions in the style of this role

Generated by Interview100x's question engine for this role. These are not questions reported from Uber.
  1. Walk me through an experiment you designed or analyzed, from the question you started with to the decision it influenced.
  2. Explain to me how you would walk a product partner through the analysis code behind a result, so they can trust the conclusion.
  3. Walk me through an experiment you designed and ran, from the hypothesis to what the result told your team.
  4. Tell me about a time product, engineering or business partners had different goals and you aligned them around an analysis or experiment result.
  5. Tell me about an analysis you owned from framing the question to influencing a decision.
  6. Tell me about a time a clean experiment was not possible and you had to choose another way to answer the question.
  7. Walk me through a metric or experiment you owned from framing the question through to the decision it influenced, including what you did and what changed as a result.
  8. Tell me about a time you persuaded partners to adopt a recommendation from your analysis that they initially doubted.
  9. Describe a time you guided teammates or another team toward a shared approach to measurement or experimentation.

A worked example

One practice question, a sample answer and what the evaluator said about it.

The question

Tell me about a time a clean experiment was not possible and you had to choose another way to answer the question.

Sample answer written for this demo

We launched a new loyalty tier in one region, and leadership wanted to know if it increased order frequency. We could not A/B test it because it was announced publicly and everyone in the region got it at the same time, so there was no control group. What I did was a difference-in-differences analysis. I picked two other regions that had not launched yet as the comparison, and I compared the change in weekly orders per active customer before and after launch in the launch region against the change in the comparison regions over the same weeks. The launch region went up more, by around four percent relative to the comparison regions, so we concluded the tier had a positive effect. I also handled a holiday that fell in the middle of the window by excluding that week. Leadership used it to approve rolling the tier out to two more regions. I liked this project because it showed you can still get a decent answer when you cannot run an experiment, as long as you are careful about what you compare against.

Interview100x's evaluator output, unedited (fill-in gaps shown in brackets)

Ready

You gave a clear, concrete account of how you measured the launch: you chose comparison regions that had not yet launched and tracked the same metric over the same weeks. The session shows a solid base, and the next step is to explain why you chose that design over alternatives and how the result shaped the decision.

  • What worked: You landed at the Ready band because you named the comparison design and the metric you used, which shows you made a deliberate choice about how to measure impact.From the answer: I picked two other regions that had not launched yet as the comparison, and I compared the change in weekly orders per active customer before and after launch in the launch region against the change in the comparison regions over the same weeks.
  • Next step: Explain why randomization was not possible in this case, such as the launch timing or the way the rollout was set up.
  • Next step: Name the alternatives you weighed, such as a before-and-after comparison within the launch region alone, and say why they fell short under the real constraints.
  • Next step: Connect the result to the decision it was meant to inform, and say how the choice held up once results came in and what you would do differently now.
  • Next step: Practice a two-minute answer on one technical decision where you explain the constraints, the alternatives you rejected, and how the result changed what happened next.
  • Note: I picked two other regions that had not launched yet as the comparison, and I compared the change in weekly orders per active customer before and after launch in the launch region against the change in the comparison regions over the same weeks. I chose this design because fill in: [why randomization was not possible]. I considered fill in: [alternatives you weighed and why they fell short]. The result was fill in: [the result and the decision it informed], and looking back I would fill in: [what you would do differently now].From the answer: I picked two other regions that had not launched yet as the comparison, and I compared the change in weekly orders per active customer before and after launch in the launch region against the change in the comparison regions over the same weeks.

What the role is evaluated on

Experimentation and statistics
A listed competency2, and Uber describes its scientists' work as experimentation, causal inference and machine learning3. Good looks like choosing the right unit of randomisation, stating what could bias the result, and knowing what you would do when a clean experiment is not possible.
Coding: SQL, Python, data structures and algorithms
All three are named under coding in the guide2. Write queries and analysis code you can explain line by line, and check edge cases such as nulls, duplicates and time zones before the interviewer asks.
Business acumen and problem solving
Both are listed2. Uber names problems such as dynamic pricing, marketplace efficiency, safety and personalisation3. Strong answers tie a metric to a decision someone at the business would actually make.
Machine learning and domain knowledge
Machine learning and domain knowledge are separate competencies in the guide2, and Uber says its scientists build and deploy models that drive real-time systems3. If your role includes them, expect to discuss how a model would be evaluated and monitored, not only how it is trained.
Behavioral and technical leadership
Both appear in the competency list2. Uber describes scientists owning metrics, experiments and solutions end to end3, so stories about owning an analysis from question to decision fit well.

Common ways answers fall short

  • Preparing for all eight competencies equally

    Uber says not every role is assessed on every competency2. Spreading a week evenly across all of them, instead of asking which apply, wastes the time you most need for the ones that do.

  • Experiment answers that ignore the marketplace

    In two-sided marketplaces generally, a change for one side can affect the other, so a simple user-level A/B test may give a misleading answer. Given that Uber lists marketplace efficiency among its science problems3, be ready to say when you would randomise by region or time instead, and what that costs you.

  • Running out of energy in the team sessions

    Four to six 45-minute sessions after earlier rounds2 is a long stretch. Answers tend to loosen late in a long run of interviews, so rehearse at least one full-length run.

  • Relying on second-hand Jam Session descriptions

    Uber's guide gives the Jam Session a name and a length, not a format2. Preparing a script based on someone else's account can leave you rigid if the session runs differently. Ask your recruiter.

  • Metrics with no decision attached

    A common weakness in data science interviews is listing metrics without saying which one decides the outcome, what threshold would change the decision, and what guardrail would stop a launch.

A prep plan for the time you have

One week

Find your competencies, then drill the top two

  1. Ask your recruiter which of the eight competencies apply to your role and which language the live coding uses2.
  2. Do three SQL problems a day with window functions and time-based joins, then explain one aloud as if to an interviewer.
  3. Write out two experiment designs for a marketplace change, each with the randomisation unit, primary metric, guardrail and the main threat to validity.
  4. Prepare one ownership story from question to decision, with the number that changed and who acted on it.
  5. Run one full voice practice session on an experimentation or product-analytics question and rework the answer the feedback rated weakest.

Suggested pass: Trial Pass, $12Use the trial on one Uber-style pricing or marketplace experiment [s9] to see whether follow-ups catch a test that could not have been run cleanly.

Two weeks

Cover coding, experimentation and business cases

  1. Days 1-4: SQL and Python practice daily; add two data structures problems if your role includes that competency2.
  2. Days 3-7: one experimentation case a day, including one where an A/B test is not possible and you need a causal inference approach.
  3. Week 2: three business cases tied to the problem areas Uber names, such as pricing, marketplace efficiency, safety or personalisation3.
  4. If machine learning applies to you, prepare to explain one model you shipped: target, features, evaluation, monitoring and what broke.
  5. Run voice practice sessions on behavioral and technical leadership questions until your stories show your own decisions.
  6. End week two with a long mock: 60 minutes of coding then four 45-minute sessions with short breaks.

Suggested pass: Sprint Pass, $99Repeated experimentation, case and behavioral rounds for one Uber Sciences role take several sessions; do the SQL and Python practice in your own editor.

Four weeks

Build depth and stamina for a long loop

  1. Week 1: baseline yourself on each competency that applies, scoring your own answers honestly.
  2. Week 1-2: statistics refresh: power and sample size, variance reduction, multiple testing and when to stop an experiment.
  3. Week 2: causal inference methods you could defend in an interview, such as difference-in-differences or synthetic control, with one worked example each.
  4. Week 3: daily business case in one of Uber's named problem areas3, answered aloud in 20 minutes.
  5. Week 3: a set of six stories covering ownership, disagreement, influence and a failed analysis.
  6. Week 4: two full-length mock days with a rest day between, then a light review.

Suggested pass: Sprint Pass, $99A month on one Uber Sciences loop, from live coding to the meet-the-team sessions [s7], is still one job; choose Search if you are interviewing for data science roles at several companies at once.

Questions candidates ask

What is the Jam Session in Uber's data science interview?
It is a 60-minute virtual stage in Uber's Sciences loop2. The guide names it and gives its length without describing the format, so ask your recruiter what it involves for your role.
How many interviews are in the Uber data scientist loop?
The Sciences guide lists a recruiter call, live coding, a Jam Session and four to six 45-minute meet-the-team sessions before the decision2.
Does Uber test SQL or Python for data scientists?
The guide lists coding, including SQL, Python and data structures and algorithms, as a possible competency2. Which ones your live coding uses depends on the role, so confirm with your recruiter.
Is the Uber applied scientist interview the same as data scientist?
Uber's Sciences guide covers data science, applied science and scientist roles together, with the same competency list, but says not every role is assessed on all competencies2. Ask which ones apply to the specific role.
Can I use AI tools during the Uber data science interview?
Uber's guides list this among their frequently asked questions2. Ask your recruiter for the current answer before the loop rather than assuming either way.

Practice with an interviewer that follows up

A Trial Pass is one full practice session with feedback on your own answers. It counts toward a Sprint Pass if you continue.

Practice this question now

Sources

  1. 1.How we hire | Uber Careers · jobs.uber.com · undated · accessed 2026-11-12
  2. 2.Sciences Interview Guide | Uber Careers · jobs.uber.com · undated · accessed 2026-11-12
  3. 3.Sciences | Uber Careers · jobs.uber.com · undated · accessed 2026-11-12

Interview100x is not affiliated with, endorsed by, or sponsored by Uber. Company names are used only to identify the employer. Process details are summarized from Uber's public pages, linked above; they can change, so check the company's own careers site.

Last reviewed November 12, 2026