Company interview guide

Lyft Machine Learning Interview Guide (2026)

Probability and marketplace cases for Data Scientists; coding, ML framing and ML design for ML Engineers.

Lyft's machine learning interview loop has 10 reported stages, from recruiter call (all tracks) to experience & values (all tracks); typical length: 3-6 weeks.

Roles covered: Data Scientist and ML Engineer. The question bank holds 20 questions reported from Lyft interviews, 2 of them free to solve in your browser.

Try a free question

Interview process length
3-6 weeks
Stages
10
Reports linked to bank questions
31
Bank report dates
Jan 2020–Jul 2026
Questions in bank
20 (2 free)
Last reviewed
October 2026

How the Lyft interview loop works

Data Scientist (both tracks): a 30-minute recruiter call, a 60-minute technical screen with a data scientist, then a virtual onsite of four or five 45-minute rounds (coding, technical case, business case, experience) on one day or split over two. ML Engineer: a recruiter call, a screen that mixes ML fundamentals with live coding (some loops open with an online coding challenge), then about four onsite rounds; one 2022 candidate was told to pass three of the four. Lyft says the data science process averages three to four weeks, yet many candidates report weeks of silence after a round. The ML Engineer outline rests on few reports, so confirm the round list with your recruiter. You can reapply after six months.

The Lyft interview process, stage by stage

01Recruiter call (all tracks)

About 30 min phone or video call

A short call on your background, why Lyft and which role or track fits. Recruiters also check work authorization, location and salary expectations; in ML Engineer loops the recruiter often lays out every later round.

What it tests

  • Whether your experience fits the Decisions track, the Algorithms track or an ML Engineer role
  • Motivation for Lyft and the team
  • Logistics: location, authorization, timeline, compensation

How to prepare

  • Be ready to rate yourself from 1 to 10 in Python, SQL and R; several recruiters ask.
  • Ask for the full list of rounds, their length and whether a take-home or online challenge is part of your loop.
  • Apply to one track and confirm the choice here; the screen can still move you to the other one.

22 candidate reports (Data Scientist, ML Engineer and one software role, a shared stage) start with a recruiter call; Lyft's data science FAQ sets it at 30 minutes with no technical content.

02Technical phone screen (Data Scientist, both tracks)

About 60 min video call with a data scientist; no live coding

One screen used for both Data Scientist tracks: probability and statistics set in rider and driver scenarios, an experiment to design and a business question to reason through. Some candidates faced two interviewers, and some were sent a Lyft blog post on how to approach the questions beforehand.

What it tests

  • Binomial and conditional probability on ride scenarios, such as whether a rider takes Lyft home
  • Expected value and launch decisions for coupons and promotions
  • Experiment design for a new feature or a driver conversion change
  • Definitions in plain words: distributions, p-values, confidence intervals, Bayesian inference
  • Diagnosing why a metric moved

How to prepare

  • Frame the problem and state assumptions before calculating; the reasoning is what gets graded.
  • Rehearse quick coupon arithmetic: revenue per ride, coupon cost, lift in rides, net revenue.
  • Read recent posts on Lyft's engineering and science blog; candidates say the questions echo them.
  • Expect follow-up on why, not only what: why this metric, why this test length.

Free reported questions for this stage

22 Data Scientist reports (15 since 2024) describe one screen on probability, statistics, experiments and Lyft business sense; Lyft's FAQ confirms 60 minutes, no coding, same for both tracks.

03Technical screen (ML Engineer)

Live call mixing ML fundamentals with a coding task; some loops open with an online coding challenge

ML Engineer loops screen on ML basics plus coding. One 2022 loop used basic ML questions and a medium practical coding problem that was neither ML nor algorithms; another 2022 loop and one from about 2025 began with an online coding challenge; a 2018 loop used a statistics and ML quiz plus a production ML design question.

What it tests

  • ML fundamentals: evaluation, overfitting, imbalanced data, model choice
  • A medium practical coding task, solved and running
  • Small ML utilities written from scratch, such as a stratified data split
  • Explaining how a model would run in production

How to prepare

  • Practise writing data-processing code quickly and cleanly; it is often not an algorithm puzzle.
  • Be able to code a stratified split or k-fold split without a library.
  • Keep ML answers short and concrete, then go deeper only when asked.

5 ML Engineer reports (four from 2018-2022, one from about 2025) describe an ML-plus-coding screen, two opening with an online challenge; Lyft's 2019 ML interview post describes the screen.

04Take-home analysis (Data Scientist, older loops)

Prompt plus anonymized data, a few days; code and a short report or slides, sometimes presented at the onsite

Many Data Scientist loops up to early 2024 included a take-home that fed a short presentation at the onsite. Lyft's 2024 description of the process leaves it out and no later report mentions it, so treat it as possible but uncommon and ask your recruiter.

What it tests

  • Exploring a messy dataset and stating assumptions
  • Turning analysis into a clear business recommendation, such as a pricing strategy
  • Effects on both sides of the marketplace

How to prepare

  • Budget well beyond the stated estimate; candidates spent a weekend to a full week on it.
  • Lead with the recommendation, keep slides short, and rehearse a 10-minute walk-through.
  • Expect questions on why you chose your method over alternatives.

8 Data Scientist reports from 2018 to early 2024 describe a multi-day data take-home; Lyft's 2021 intern guide includes it, but its 2024 FAQ does not and no later report mentions it.

05Onsite: coding (Data Scientist)

45 min live coding in your chosen language; Algorithms track: Python algorithms; Decisions track: analytical SQL or Python on a rideshare dataset

Algorithms-track candidates write a working end-to-end solution to an algorithm, ML implementation or data-processing task, sometimes at medium difficulty. Decisions-track candidates answer a series of analytical questions on a sample rideshare dataset, usually in SQL. One Toronto loop in 2025 had both an algorithm and a decisions coding round.

What it tests

  • Clean, working code from start to finish within the slot
  • Standard data structures: heaps, deques, hash maps, prefix sums
  • Sampling and small ML utilities written from scratch
  • SQL joins, aggregation and window functions on trip data (Decisions track)

How to prepare

  • Discuss the approach and pitfalls before typing; Lyft says that discussion is graded.
  • Expect practical problems rather than tricks, but practise medium-level array and sampling problems.
  • For the Decisions track, drill SQL on rides, drivers and time windows until it is routine.

Free reported questions for this stage

8 Data Scientist reports name an onsite coding round (SQL, Python or a medium algorithm problem); Lyft's FAQ describes the split between the Algorithms and Decisions versions.

06Onsite: coding rounds (ML Engineer)

Usually two rounds in a shared IDE or on a laptop: an algorithms problem and a practical build

ML Engineer loops include an algorithms round, reported as medium level, and a practical round: an object-oriented build on a laptop, an end-to-end coding task, or an ML algorithm such as k-means written from scratch. A 2018 candidate said questions were easy to medium but graded strictly on correctness and efficiency.

What it tests

  • Medium algorithm problems, such as interval overlap
  • Object-oriented design of a small working module
  • Implementing a classic ML algorithm such as k-means without libraries
  • Correct, efficient code with edge cases handled

How to prepare

  • Practise building a small multi-class module from a spec and running it on sample input.
  • Know k-means, k-nearest neighbours and logistic regression well enough to code them in 20 minutes.
  • Aim for fully working solutions; one candidate who thought they did well was rejected for not finishing.

4 ML Engineer reports (2018-2022 and about 2025) describe an algorithms round plus a practical or object-oriented build; Lyft's general engineering hiring page describes a similar laptop task for software engineers.

07Onsite: technical or ML case (all tracks, content differs)

45 min open discussion for Data Scientists; ML Engineer length not reported

Algorithms-track and ML Engineer candidates get a loosely described Lyft problem and must turn it into a defined modeling or optimization task, then say how to evaluate it. Decisions-track candidates diagnose a product problem end to end with experiments, probability and product sense. Lyft's ML interview post says the candidate is expected to lead and that overly elaborate answers get steered back to simple heuristics.

What it tests

  • Turning a vague business problem into regression, classification or optimization
  • Choosing offline and online metrics and judging whether the data is enough
  • Diagnosing why a marketplace metric changed
  • Experiments when riders or drivers share a pool: interference, switchbacks, geo splits

How to prepare

  • Drive the conversation; interviewers deliberately let you lead.
  • Start with a simple baseline or heuristic before proposing anything elaborate.
  • Always ask what happens to the other side of the marketplace.

8 reports (5 Data Scientist, 3 ML Engineer) describe an ML, optimization or product case round; Lyft's data science FAQ and its 2019 ML interview post both describe it.

08ML system design (ML Engineer, most loops)

Design discussion; some loops add a separate large-scale system design round

ML Engineer candidates design an ML-backed product end to end. Reported prompts include a text completion system and recommending nearby attractions to riders who open the app; a 2018 loop had both an ML design call and a large-scale system design round. Recent third-party guides also list a dedicated ML design round, but first-hand detail is thin.

What it tests

  • Scoping the product goal and the prediction target
  • Data, features and labels from rides, locations and user history
  • Serving with low latency and keeping training and serving features consistent
  • Offline metrics, an online A/B test and monitoring after launch

How to prepare

  • Prepare one recommender design and one real-time prediction design (such as ETA) in Lyft terms.
  • Agree on scope and success metrics in the first few minutes.
  • Say how you would measure each part before going deep on any of them.

3 ML Engineer reports (2018, 2021, about 2025) include an ML system design prompt; third-party prep guides also list the round, but no Lyft page describes it.

09Onsite: business case (Data Scientist)

45 min, no coding

A business problem built on Lyft metrics, such as whether a promotion pays off. Candidates describe a structured answer with clear metrics and trade-offs as the bar. In older loops some Decisions-track candidates presented their take-home instead.

What it tests

  • Structuring an open business question
  • Unit economics: fares, driver pay, contribution margin
  • Short-term revenue against long-term growth and supply effects
  • Making a concrete recommendation

How to prepare

  • Write the per-ride economics down early and keep them visible.
  • Finish with a decision and the one number that would change it.

6 Data Scientist reports (2022-2026) name a business case round; Lyft's data science FAQ describes it as a no-coding problem built on Lyft metrics.

10Experience & values (all tracks)

About 45 min behavioral and past-project interview with a data scientist, manager or HR

A deep dive into past projects and how you work. Lyft's guides name end-to-end ownership, analysis that is rigorous but pragmatic, empathy and integrity, and collaboration as what they look for. ML Engineer reports call it the experience or values round.

What it tests

  • Ownership of a project from question to shipped result
  • A mistake you made and what you learned
  • Collaboration and handling disagreement
  • Concise communication to non-technical partners

How to prepare

  • Prepare two or three project stories with the metric you moved and your exact role.
  • Keep answers short; one rejected candidate was told communication was not concise enough.
  • Have a story ready on driving change in a team that resisted it.

10 reports across tracks (one from a software role, a shared stage) describe a behavioral and past-project round; Lyft's data science FAQ and intern guide describe it and list the values.

Questions and themes reported most often

  • Take-home analysis on a Lyft dataset, then present it (8 reports, take-home analysis (data scientist, older loops))
  • Design an experiment for a new feature or a driver-side change (5 reports, technical phone screen (data scientist, both tracks))
  • Binomial and conditional probability set in a ride scenario (4 reports, technical phone screen (data scientist, both tracks))
  • SQL on trip and driver data (4 reports, onsite: coding (data scientist))
  • Should a rider coupon or promotion launch? Expected value and net revenue (3 reports, technical phone screen (data scientist, both tracks))
  • Define p-values, confidence intervals and distributions in plain words (3 reports, technical phone screen (data scientist, both tracks))
  • Bayes' theorem applied to a business question (3 reports, technical phone screen (data scientist, both tracks))
  • Classic probability puzzles: coin patterns, dice races, cards (3 reports, technical phone screen (data scientist, both tracks))

What Lyft asks about

  • Probability & statistics25%
  • Coding (Python, SQL, algorithms)22%
  • Experimentation & product metrics20%
  • ML case & ML system design18%
  • Business case7%
  • Behavioral & experience8%

Lyft interview prep checklist

  • Can decide whether a coupon should launch from per-ride revenue, coupon cost and lift
  • Solved a dozen probability puzzles (patterns, races, conditional counts) out loud
  • Can design a driver-side experiment and explain why rider-level randomization would be biased
  • Can explain a p-value and a confidence interval to a non-technical partner
  • Wrote a weighted sampler, a stratified split and k-means from scratch in under 20 minutes each
  • Can turn a vague Lyft problem into an ML task with a baseline, metrics and a serving plan
  • Prepared two project stories with the metric you moved and what you owned

Tips

  • Decide your track early: Decisions leans on experiments, SQL and product sense; Algorithms on Python, modeling and optimization; ML Engineer on coding, ML framing and system design.
  • Learn the basics of the business: riders, drivers, fares, driver pay, incentives, wait times and ETAs; most questions assume them.
  • Drill business-flavoured probability until binomial, conditional and expected-value questions take a minute or two.
  • Always check for marketplace interference when designing an experiment, and know geo, time-sliced and switchback alternatives.
  • In every case round, structure first: goal, metric, approach, risks, recommendation, and lead the discussion yourself.
  • ML Engineer candidates should expect a strict bar on coding: practise medium algorithm problems and a small object-oriented build to fully working code.
  • Practise saying technical ideas plainly; concise communication is graded in every round.
  • Read recent posts on Lyft's engineering and science blog; candidates say themes overlap with interview questions.
  • Plan around delays and follow up politely; long silences after a round are the most common complaint.

See the rest of the Lyft loop

The later stages, the questions reported for them and how to prepare are in Pro.

Unlock with Pro

Practice the Lyft loop

Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.

Lyft interview FAQ

How many interview rounds does Lyft have for ML roles?

Candidates report 10 stages: Recruiter call (all tracks), Technical phone screen (Data Scientist, both tracks), Technical screen (ML Engineer), Take-home analysis (Data Scientist, older loops), Onsite: coding (Data Scientist), Onsite: coding rounds (ML Engineer), Onsite: technical or ML case (all tracks, content differs), ML system design (ML Engineer, most loops), Onsite: business case (Data Scientist) and Experience & values (all tracks). Data Scientist (both tracks): a 30-minute recruiter call, a 60-minute technical screen with a data scientist, then a virtual onsite of four or five 45-minute rounds (coding, technical case, business case, experience) on one day or split over two. ML Engineer: a recruiter call, a screen that mixes ML fundamentals with live coding (some loops open with an online coding challenge), then about four onsite rounds; one 2022 candidate was told to pass three of the four. Lyft says the data science process averages three to four weeks, yet many candidates report weeks of silence after a round. The ML Engineer outline rests on few reports, so confirm the round list with your recruiter. You can reapply after six months.

How long does the Lyft ML interview process take?

Reported timelines run 3-6 weeks.

What does Lyft ask in ML interviews?

By reported question mix: Probability & statistics 25%; Coding (Python, SQL, algorithms) 22%; Experimentation & product metrics 20%; ML case & ML system design 18%; Business case 7%; Behavioral & experience 8%. The most repeated themes: Take-home analysis on a Lyft dataset, then present it; Design an experiment for a new feature or a driver-side change; Binomial and conditional probability set in a ride scenario.

How should I prepare for the Lyft ML interview?

Decide your track early: Decisions leans on experiments, SQL and product sense; Algorithms on Python, modeling and optimization; ML Engineer on coding, ML framing and system design. Learn the basics of the business: riders, drivers, fares, driver pay, incentives, wait times and ETAs; most questions assume them. Drill business-flavoured probability until binomial, conditional and expected-value questions take a minute or two.

Are there free Lyft interview questions?

Yes. LeetML has 20 questions reported from Lyft interviews, and 2 of them are free to solve in the browser; they are listed by stage above.

Related interview guides