Company interview guide
X (formerly Twitter) Machine Learning Interview Guide (2026)
Algorithm coding, ML or stats from first principles, a design or case round and a close look at your past work.
X (formerly Twitter)'s machine learning interview loop has 7 reported stages, from recruiter or hiring-manager intro to behavioral, project deep dive and hiring manager; typical length: About 3-4 weeks in the few reports that say.
Roles covered: ML Engineer, Applied Scientist and Data Scientist. The question bank holds 7 questions reported from X (formerly Twitter) interviews.
Try a free question from this loop's topics
Part of SpaceX. Twitter was renamed X in 2023. X merged into xAI in March 2025, and xAI merged into SpaceX in February 2026. Questions here include reports from the Twitter era, so expect the current loop to differ.
- Interview process length
- About 3-4 weeks in the few reports that say
- Stages
- 7
- Reports linked to bank questions
- 9
- Bank report dates
- Jun 2020–Dec 2023
- Questions in bank
- 7 (0 free)
- Last reviewed
- October 2026
How the X (formerly Twitter) interview loop works
In Twitter-era reports the loop ran: a recruiter or hiring-manager intro call, sometimes an online assessment or take-home (data science teams), one technical screen of 30-60 minutes, then a final round of three to five sessions on one day. Final sessions often had two interviewers. ML engineering finals mixed algorithm coding, ML breadth, an ML system design or case study, a deep dive on a past project and a behavioral or hiring-manager conversation; data science finals swapped design for probability, experimentation and product cases. Several candidates were told about format changes late, so confirm every round with the recruiter. Reports since X joined xAI are too few to describe the current loop; the xAI careers page lists an application review, a short screening call, in-depth technical interviews and an offer, with technical staff rather than recruiters doing the assessing.
The X (formerly Twitter) interview process, stage by stage
01Recruiter or hiring-manager intro
15-60 min call; often with the hiring manager or a team lead rather than only a recruiter
A first conversation about your background, motivation and the team. In about two thirds of the reports the hiring manager, an engineering manager or a team lead ran it, and several of those calls included technical or product questions about your past work, so treat it as an interview rather than logistics.
What it tests
- A clear story of your ML or data work and what you personally built
- Why this team and product
- Product sense for data science roles, for example how you would measure new-user engagement
- Logistics: location, office expectations, level
How to prepare
- Prepare a two-minute summary of one project with numbers, then be ready to go deeper on it.
- Ask which rounds follow and what each covers; several candidates were surprised by late format changes.
- Expect a direct question about working from the office full time in recent reports.
Intro call in 15 ML and data science reports (mostly 2018-2022) plus one 2025 engineering report; ten were with a manager or team lead. The xAI careers page, where X's careers link now goes, lists a short screening call.
02Online assessment or take-home (some data science teams)
An online coding test, or a take-home with a probability problem and a business case; not reported for ML engineer roles
Some data science candidates had an automated coding test or a take-home before the technical screen or the final round. Take-homes combined a multi-part probability question with an A/B testing or business case, and one final round included a presentation. No ML engineer report mentions this step.
What it tests
- Probability with several sub-questions
- Designing and reading an A/B test for a product change
- Turning a business question into an analysis plan
- Basic coding under time pressure
How to prepare
- Write the take-home so it can be presented: assumptions, method, result, recommendation.
- Show the arithmetic of every probability step; partial credit depends on it.
Seven data science reports from 2018-2022: four take-homes (probability plus a business or A/B case) and three online coding tests. No ML engineer report has one.
03Technical screen
One 30-60 min video or phone round; ML roles mostly algorithm coding with an engineer, data science roles coding, SQL, stats or a product question
Usually a single round. For ML engineers it was most often a standard algorithm problem with an optimisation follow-up and a complexity discussion, after about ten minutes on the resume; a few got ML problem solving instead of the coding they were told to expect. Data science screens mixed short Python data manipulation (sometimes without built-in functions), SQL, hypothesis testing and product metrics.
What it tests
- Strings, arrays, matrix traversal with BFS or DFS and a trie follow-up
- Optimising a first solution and stating time and space complexity
- Counting events per user over time windows of different sizes
- SQL and Python data manipulation for data science roles
- Framing a hypothesis test or a product metric
How to prepare
- Talk through the idea before coding; interviewers asked for an optimised version once the basic one worked.
- Drill medium graph and string problems, including a trie follow-up.
- For data science screens, practise grouping and windowed counts by hand in plain Python as well as in SQL.
A single technical screen in 9 ML and 11 data science reports, 2015-2024; ML screens were mostly algorithm coding, and two candidates got ML questions instead of the coding they were told about.
04Final round: coding and ML from scratch
1-2 sessions of about 45-60 min, often with two interviewers
Final rounds kept at least one coding session: algorithm problems for most candidates, sometimes harder than the screen. ML and data science candidates were also asked to implement a classic algorithm from scratch, such as k-means, k-nearest neighbours or a decision tree, and one candidate was warned to expect an ML coding question at the onsite.
What it tests
- Medium to hard algorithm problems
- Implementing k-means, kNN or a decision tree without libraries
- Python library fundamentals for data roles
- Explaining the cost and convergence of what you wrote
How to prepare
- Be able to write k-means, kNN and an entropy-based decision tree split in plain Python or numpy in under 30 minutes.
- With two interviewers, keep both in the loop: state the plan out loud before you type.
- Do not skip algorithm practice for ML roles; recruiters sometimes changed the mix late.
Coding or ML-from-scratch work after the screen in 12 ML and data science reports, 2015-2023, including two candidates' own loop schedules. The xAI careers page lists in-depth technical interviews.
05Final round: ML breadth and ML system design
1-2 sessions in ML engineer loops: an ML breadth round, often with a small case, and a larger design or case-study discussion
ML engineering finals usually had one round on ML fundamentals and one on designing a model or system for a product problem, sometimes run as a big case study. Prompts came from the product, such as how to represent a post for a model, and one data science loop had a distributed computing design round instead. One candidate found the problem poorly posed, so drive the scoping yourself.
What it tests
- ML fundamentals: bias and variance, losses, evaluation, regularisation
- Framing a product problem as an ML task with labels and metrics
- Feature and representation choices for short text and user activity
- Serving and scaling trade-offs
- Applied retrieval and search ideas
How to prepare
- Prepare a feed or timeline ranking design and a harmful-content detection design; both fit the product.
- Ask clarifying questions until the goal and metric are explicit, then propose a baseline first.
- Review classic ML theory; breadth rounds went quickly from definitions to trade-offs.
ML breadth or ML design in 6 ML-track reports, 2015-2021, one of them a candidate's own loop schedule; a 2023 data science loop had a distributed systems design round instead. Prompts are rarely given.
06Data science: probability, experimentation and product cases
Data science roles: 1-2 final sessions plus parts of the screen; probability, A/B testing, metrics and causal questions
The core of data science loops. Expect A/B testing in depth, often for a feature on the platform, product metrics for engagement and retention, probability puzzles and sometimes a causal-inference case with sample code on user data. One senior candidate who was told A/B experience was not required was still asked mostly A/B questions.
What it tests
- Designing, sizing and reading an A/B test, including pitfalls
- Choosing KPIs for new-user activation and retention
- Probability and expected value puzzles
- Causal reasoning when a clean experiment is not possible
- Market sizing and product judgement
How to prepare
- Prepare A/B testing end to end regardless of what the recruiter says about experience.
- Practise network-effect and interference issues in experiments on a social product.
- Solve probability puzzles with a symmetry or indicator argument and sanity-check the answer.
Probability, A/B testing or product-metric questions in 14 data science reports, 2018-2024, in screens, take-homes and finals; A/B testing appears in five.
07Behavioral, project deep dive and hiring manager
1-2 sessions of 45-60 min: behavioral questions, a deep dive on a past project, and a hiring-manager conversation; one ML loop called a round topgrading
Interviewers dug into prior ML work in detail, and in one ML loop a structured career-history round was scheduled. Several data science finals ended with the hiring manager, and one included a presentation. The most recent ML report adds a direct question about being in the office at all times.
What it tests
- Depth and ownership behind each project on your resume
- Decisions you made, alternatives and results
- Collaboration and handling disagreement
- Commitment to in-office work and the pace of the team
How to prepare
- Prepare a chronological walk through every role: goals, results, why you left.
- Pick two projects and rehearse the follow-ups on scale, trade-offs and your own contribution.
- Decide before the loop whether full-time office work and a possible relocation suit you.
Behavioral, deep-dive or hiring-manager rounds in 11 ML and data science reports, 2020-2024, plus one 2025 engineering report; questions on past projects recur.
Questions and themes reported most often
- Standard algorithm problem with an optimisation follow-up (strings, matrix BFS or DFS, top-n) (6 reports, technical screen)
- Deep dive on a past ML or data project (6 reports, behavioral, project deep dive and hiring manager)
- Design or analyse an A/B test for a platform feature (5 reports, data science: probability, experimentation and product cases)
- SQL questions in the data science screen (5 reports, technical screen)
- ML case study or system design for a product problem (4 reports, final round: ml breadth and ml system design)
- Probability puzzle with dice or a distribution (4 reports, data science: probability, experimentation and product cases)
- Implement k-means, kNN or a decision tree from scratch (3 reports, final round: coding and ml from scratch)
- Count user actions over time windows of different granularity (3 reports, technical screen)
What X (formerly Twitter) asks about
- Algorithm coding33%
- Stats, experimentation & product20%
- ML system design & case studies15%
- ML coding from scratch10%
- ML breadth10%
- Behavioral & project deep dive12%
X (formerly Twitter) interview prep checklist
- Solved medium graph, string and trie problems with a spoken complexity analysis
- Wrote k-means, kNN and an entropy-based decision tree from scratch
- Can design and critique an A/B test for a feature on a social product
- Can pick and defend KPIs for new-user activation and retention
- Sketched a feed-ranking ML design with labels, features, metrics and serving
- Prepared a deep dive on two projects and a walk through your career history
- Confirmed with the recruiter the current rounds, office expectations and location
Tips
- Read this page as a guide to the Twitter-era loop; ask your recruiter how the current process differs before you plan.
- Make algorithm coding your first priority; it ran through the screen and the final round for ML roles.
- Be ready to write k-means, kNN and a decision tree from scratch without libraries.
- For data science roles, prepare A/B testing in depth, including experiments on a network of users.
- Prepare one feed-ranking and one content-safety ML design, scoped from a vague product prompt.
- Rehearse a detailed walk through your past projects and roles; deep dives were common.
- Get every round's type confirmed in writing; several candidates heard about format changes late.
- Follow up politely if you hear nothing; slow or missing recruiter replies come up often in reports.
Practice the X (formerly Twitter) loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
X (formerly Twitter) interview FAQ
How many interview rounds does X (formerly Twitter) have for ML roles?
Candidates report 7 stages: Recruiter or hiring-manager intro, Online assessment or take-home (some data science teams), Technical screen, Final round: coding and ML from scratch, Final round: ML breadth and ML system design, Data science: probability, experimentation and product cases and Behavioral, project deep dive and hiring manager. In Twitter-era reports the loop ran: a recruiter or hiring-manager intro call, sometimes an online assessment or take-home (data science teams), one technical screen of 30-60 minutes, then a final round of three to five sessions on one day. Final sessions often had two interviewers. ML engineering finals mixed algorithm coding, ML breadth, an ML system design or case study, a deep dive on a past project and a behavioral or hiring-manager conversation; data science finals swapped design for probability, experimentation and product cases. Several candidates were told about format changes late, so confirm every round with the recruiter. Reports since X joined xAI are too few to describe the current loop; the xAI careers page lists an application review, a short screening call, in-depth technical interviews and an offer, with technical staff rather than recruiters doing the assessing.
How long does the X (formerly Twitter) ML interview process take?
Reported timelines run About 3-4 weeks in the few reports that say.
What does X (formerly Twitter) ask in ML interviews?
By reported question mix: Algorithm coding 33%; Stats, experimentation & product 20%; ML system design & case studies 15%; ML coding from scratch 10%; ML breadth 10%; Behavioral & project deep dive 12%. The most repeated themes: Standard algorithm problem with an optimisation follow-up (strings, matrix BFS or DFS, top-n); Deep dive on a past ML or data project; Design or analyse an A/B test for a platform feature.
How should I prepare for the X (formerly Twitter) ML interview?
Read this page as a guide to the Twitter-era loop; ask your recruiter how the current process differs before you plan. Make algorithm coding your first priority; it ran through the screen and the final round for ML roles. Be ready to write k-means, kNN and a decision tree from scratch without libraries.
Are there free X (formerly Twitter) interview questions?
LeetML has 7 questions reported from X (formerly Twitter) interviews. The stage sections above link the free practice problems for the topics each round covers; the reported-question set itself is in the app.
Related interview guides
- xAIA fast loop run by engineers: a short screen, production-style coding, design, and a deep dive on your own work.
- MetaFast classic coding, an AI-assisted coding project and product ML design for recommendations and integrity.
- LinkedInProbability-flavoured coding, ML fundamentals and recommender design built around LinkedIn's own products.
- SnapClassic coding plus fast, broad ML fundamentals and recommendation-heavy ML design, with values questions in every round.
- PinterestRapid-fire ML fundamentals, edge-case-heavy coding and embedding-centric recommendation design.
- RedditA live model-building screen on a small dataset, then practical coding, ML fundamentals, ranking-system design and a hiring manager round.