
Company interview guide
Citi Machine Learning Interview Guide (2026)
A thinly documented loop: an application and skills check, a background interview, then a final stage of technical interviews, with quant roles leaning on probability and markets maths.
Citi's machine learning interview loop has 6 reported stages (Application & skills evaluation, First interview, Coding questions, Probability & statistics, Markets and quant finance (quant track), Final stage & behavioral); typical length: Varies by program; not well reported.
Roles covered: Data Scientist, ML Engineer and Research Scientist. The question bank holds 7 questions reported from Citi interviews, 5 of them free to solve in your browser.
- Interview process length
- Varies by program; not well reported
- Stages
- 6
- Reports linked to bank questions
- 7
- Bank report dates
- Jun 2017–Jun 2026
- Questions in bank
- 7 (5 free)
- Last reviewed
- October 2026
How the Citi interview loop works
Citi's early-career pages describe five steps: an online application, a skills evaluation (a coding assessment for technology programs, a skills survey for some EMEA and Services programs), a one-on-one first interview about your background and goals, a final stage that can be several interviews, an assessment or a case study depending on program and location, and then an offer. Analyst program applications usually open in September of the year before the start date. The bank does not publish a process for experienced hires, and we have no first-hand reports for quant, data science or ML roles, so the round count, durations and the split between coding, probability and markets questions are unconfirmed. The question pool we hold suggests quant interviews test probability and statistics, while coding rounds use standard algorithm problems.
The Citi interview process, stage by stage
01Application & skills evaluation
Online application, then an online skills evaluation: a coding assessment for technology programs or a skills survey for some EMEA and Services programs
Citi's early-career pages put a skills evaluation straight after the application. For technology programs it is a coding assessment; for some EMEA and Services programs it is a skills discovery survey instead. Content, length and platform are not described for quant, data or ML roles, and experienced-hire postings may skip this step.
What it tests
- Basic algorithm and data-structure fluency under time pressure (technology programs)
- How your skills and interests map to the program you applied to
- A resume that clearly shows quantitative and programming experience
How to prepare
- Use a new Workday account created for Citi; the bank says accounts from other employers cannot be reused.
- Highlight concrete quantitative, modelling or coding work on the application, since that is what the first screen sees.
- Practise timed array, dynamic programming and greedy problems in case the assessment is coding-based.
Described on the bank's early-careers page (coding assessment for technology programs); no first-hand reports for quant, data or ML roles confirm its content or length.
02First interview
One-on-one interview on your background and how your goals fit the bank
The bank describes a single first interview focused on your experience and why you want the role. For quant and data roles expect the interviewer to probe the projects on your resume, and possibly to ask a first technical or probability question, although we have no reports confirming that.
What it tests
- A clear story of your background and the role you want
- Why Citi, and why this desk or team
- Depth behind the quantitative or ML projects on your resume
How to prepare
- Prepare a two-minute walkthrough of one project with the data, the method and the result.
- Know what the target group does (for example Markets quants versus risk or analytics) and say why it fits you.
- Ask the interviewer which rounds follow and what each one covers.
A one-on-one first interview on background and fit is described on the bank's early-careers page; no first-hand reports for quant, data or ML roles.
03Coding questions
Live or online algorithm problems; format and duration not reported for quant, data or ML roles
The coding questions tagged to Citi come from software engineering interviews between 2020 and 2026: a maximum circular subarray sum, the longest increasing subsequence in n log n time and interval scheduling on one or several machines. They are classic dynamic programming and greedy problems rather than ML coding. Quant developer and data roles are likely to see similar problems, but that is not confirmed.
What it tests
- Linear-time dynamic programming over arrays, including wrap-around cases
- Patience-sorting style n log n algorithms and reconstructing the answer
- Greedy scheduling and proving why the greedy choice is optimal
- Stating time and space complexity precisely
How to prepare
- Explain the invariant behind each DP or greedy step before you code it.
- Handle edge cases out loud: all-negative arrays, ties, touching intervals.
- Write in the language you are fastest in; nothing we hold says the bank mandates one.
Free reported questions for this stage
- Maximum Sum Circular SubarrayCoding & AlgorithmsMediumSolve
- Longest Increasing SubsequenceCoding & AlgorithmsHardSolve
- Maximum Non-Overlapping Interval SchedulingCoding & AlgorithmsHardSolve
Only three coding problems tagged to software engineering interviews at the bank (2020-2026); no first-hand reports describe a coding round for quant, data or ML roles.
04Probability & statistics
Whiteboard or verbal problems; format and duration not reported
Quant researcher and quant trader interviews tagged to Citi ask short probability and statistics questions: a Bayes problem with boxes of balls, uniform order statistics, the number of records in a random permutation and the tightest tail bound from two moments. Data science and ML candidates should expect the same family of questions plus core statistics, but we have no reports from those roles.
What it tests
- Conditional probability and Bayes' rule with careful conditioning
- Expectation and variance via indicator variables
- Order statistics of uniform samples and their limits
- Moment inequalities and constructing distributions that make a bound tight
How to prepare
- Reach for indicators and linearity of expectation before brute-force counting.
- For bound questions, name the inequality first, then show a distribution that attains it.
- Talk through the reasoning; quick sanity checks on small cases earn credit.
Free reported questions for this stage
- Three Boxes: Conditional Probability of a White BallProbabilityEasySolve
- Expected Number and Variance of Records in a Random PermutationProbabilityMediumSolve
Four probability problems tagged to quant researcher and trader interviews at the bank (2017-2026); compiled question lists agree probability is a core block. No first-hand round descriptions.
05Markets and quant finance (quant track)
Technical discussion with a desk quant or strategist; Markets quant roles only, unconfirmed
Compiled question lists for Citi quant and strats roles lean heavily on fixed income and derivatives: pricing caps, swaptions and FX products, curve building, DV01 and key-rate hedging, and covariance work such as PCA of rates or repairing a covariance matrix. None of these is in our practice pool for Citi yet, and data science or ML roles outside Markets are unlikely to get them.
What it tests
- No-arbitrage pricing, put-call parity and risk-neutral valuation
- Rates sensitivities: duration, convexity, DV01 and key-rate hedges
- Covariance estimation, PCA and why the first rates factor is a level shift
- Time-series basics such as ARMA and GARCH stationarity
How to prepare
- Tie every hedge to a sensitivity and a no-arbitrage argument.
- Be ready to bootstrap discount factors from par swap rates by hand.
- Ask early whether your loop includes a desk round; analytics and ML roles often do not.
Rests only on compiled question lists for the bank's quant and strats roles; no first-hand reports, and nothing on whether data or ML roles include it.
06Final stage & behavioral
Depending on program and location: several interviews, an assessment or a case study exercise
The bank says the final stage varies by program and location and can combine several interviews, an assessment or a case study. Expect a mix of technical follow-ups and questions on motivation and teamwork with people from the hiring group. How many rounds, how long and what the case study covers for quant, data or ML roles is not reported.
What it tests
- Teamwork, communication and handling pressure
- Structuring an open-ended business or data case
- Consistency with the story you told in the first interview
- Motivation for the specific group and location
How to prepare
- Prepare three stories (a hard problem, a disagreement, a mistake) with your own contribution clear.
- If a case study is announced, ask whether it is a markets, data or business case and whether you can use code.
- Have questions ready about the team's models, data and tooling.
The bank's early-careers page says the final stage can be several interviews, an assessment or a case study; no first-hand reports for quant, data or ML roles add detail.
Questions and themes reported most often
- Tightest bound on a tail probability from the first two moments (1 reports, probability & statistics)
- Mean and variance of uniform order statistics (1 reports, probability & statistics)
- Conditional probability with three boxes of coloured balls (1 reports, probability & statistics)
- Expected number and variance of records in a random permutation (1 reports, probability & statistics)
- Maximum sum circular subarray in linear time (1 reports, coding questions)
- Longest increasing subsequence in n log n with reconstruction (1 reports, coding questions)
- Maximum non-overlapping interval scheduling on one or more machines (1 reports, coding questions)
What Citi asks about
- Probability & statistics30%
- Coding & algorithms25%
- Rates, derivatives & quant finance (quant track)20%
- Linear algebra & time series10%
- Background, motivation & behavioral15%
Citi interview prep checklist
- Confirmed with the recruiter the number, type and length of every round
- Solved the circular subarray, LIS and interval scheduling problems with complexity stated
- Worked the order statistics, records and moment-bound probability problems without notes
- Can explain Bayes' rule and indicator-variable arguments out loud on a fresh problem
- Reviewed rates and derivatives basics if the role sits in Markets
- Prepared a two-minute project walkthrough and three behavioral stories
- Know why this specific group at Citi, and what it builds
Tips
- Ask the recruiter which group the role sits in (Markets quants, risk, technology or analytics); the loop follows the group.
- Drill short probability problems: Bayes, indicators, order statistics and moment bounds.
- Keep classic array DP and greedy problems sharp for any coding assessment or live round.
- For Markets quant roles, review rates and derivatives basics: discounting, DV01, Black's model and put-call parity.
- Refresh linear algebra used in risk: covariance matrices, PCA and positive semi-definiteness.
- For data science and ML roles, be ready to explain a model you built end to end, including data issues and validation.
- Apply early for analyst programs; the bank says applications usually open in September of the year before the start.
- Treat this outline as provisional; it rests on the bank's own process page and a handful of tagged questions, not on candidate reports.
See the rest of the Citi loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Citi loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Citi interview FAQ
How many interview rounds does Citi have for ML roles?
Candidates report 6 stages: Application & skills evaluation, First interview, Coding questions, Probability & statistics, Markets and quant finance (quant track) and Final stage & behavioral. Citi's early-career pages describe five steps: an online application, a skills evaluation (a coding assessment for technology programs, a skills survey for some EMEA and Services programs), a one-on-one first interview about your background and goals, a final stage that can be several interviews, an assessment or a case study depending on program and location, and then an offer. Analyst program applications usually open in September of the year before the start date. The bank does not publish a process for experienced hires, and we have no first-hand reports for quant, data science or ML roles, so the round count, durations and the split between coding, probability and markets questions are unconfirmed. The question pool we hold suggests quant interviews test probability and statistics, while coding rounds use standard algorithm problems.
How long does the Citi ML interview process take?
Reported timelines run Varies by program; not well reported.
What does Citi ask in ML interviews?
By reported question mix: Probability & statistics 30%; Coding & algorithms 25%; Rates, derivatives & quant finance (quant track) 20%; Linear algebra & time series 10%; Background, motivation & behavioral 15%. The most repeated themes: Tightest bound on a tail probability from the first two moments; Mean and variance of uniform order statistics; Conditional probability with three boxes of coloured balls.
How should I prepare for the Citi ML interview?
Ask the recruiter which group the role sits in (Markets quants, risk, technology or analytics); the loop follows the group. Drill short probability problems: Bayes, indicators, order statistics and moment bounds. Keep classic array DP and greedy problems sharp for any coding assessment or live round.
Are there free Citi interview questions?
Yes. LeetML has 7 questions reported from Citi interviews, and 5 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- JPMorgan ChaseA recorded first-round assessment, then team interviews that test ML basics, your own projects and, for quant tracks, probability and statistics.
- Bank of AmericaA conventional bank loop: recruiter and hiring-manager calls, a technical conversation, sometimes a take-home, with statistics-heavy questions for quant roles.
- BarclaysShort online tests, then experience and technical interviews for quant, data science and ML roles, with a strong values check.
- Two SigmaInterview guide for ML, AI and research roles: a numeric HackerRank assessment, then data science, coding and regression-heavy statistics rounds, three manager rounds and a senior closer.