Company interview guide
IMC Trading Machine Learning Interview Guide (2026)
For ML, AI and research roles at a global options market maker: a track-split timed test, probability and ML-fundamentals rounds, and a fast superday.
IMC Trading's machine learning interview loop has 7 reported stages, from hr screen to superday / final rounds; typical length: Fast once it starts: compiled accounts describe superdays booked within days of the first round, though seats are often frozen or filled.
Roles covered: ML Engineer, Research Scientist, Deep Learning Engineer and Applied Scientist. The question bank holds 6 questions reported from IMC Trading interviews, 3 of them free to solve in your browser.
- Interview process length
- Fast once it starts: compiled accounts describe superdays booked within days of the first round, though seats are often frozen or filled
- Stages
- 7
- Reports linked to bank questions
- 6
- Bank report dates
- Nov 2024–Jun 2026
- Questions in bank
- 6 (3 free)
- Last reviewed
- October 2026
How the IMC Trading interview loop works
An application (the firm asks you to pick one role and uses an AI tool to organize resumes in some locations), an HR screen of 10-45 minutes before or soon after the test, a timed online test that splits by track (quant research: 11 math, statistics and probability questions, or in older accounts 15 questions in 60 minutes; developer and ML engineer: 120 minutes for two hard algorithm problems), a one-way recorded video scored as part of that test, then a 45-60 minute technical round and a superday of about five rounds on site or virtual. Engineering candidates get two-hour live coding sessions with a systems deep dive, or for some Python roles a 3-5 hour take-home. The market-making and betting game, the mental-math and number-sequence tests and the 2026 pattern-plus-personality screen are trader-track only. With two first-hand ML reports, treat this as the quant research loop adapted for ML titles.
The IMC Trading interview process, stage by stage
01HR screen
10-45 min phone or video call with HR, before or a few days after the online test; no technical questions in recent research-track accounts
Admin first: graduation date, visa and sponsorship, notice period, expected salary and work mode. Then why the firm, why research or trading, and a walk through two resume projects end to end. Behavioral prompts cover a hard-to-hear opinion, an advisor who changed your approach, conflict, failure and other offers. Research candidates are asked how machine learning makes money in finance, with concrete results. Research-track candidates hear back in about one to two weeks.
What it tests
- A specific, personal answer to why the firm and why trading
- Two research or ML projects explained end to end with a number attached
- How ML would create value in trading: deep learning, Sharpe, P&L
- Logistics: sponsorship, start date, competing processes
How to prepare
- Attach a measurable outcome to each project: accuracy, Sharpe, latency saved or P&L, and say what the baseline was.
- Give the same why-the-firm story you will repeat later; it is asked again at the superday.
- Tell the recruiter about competing offer deadlines early; the firm's own page asks you to.
- Have a one-line answer on visa needs; offices differ on sponsorship.
One 2022 ML researcher report starts with an HR screen. Content and timing follow compiled accounts of the quant research and trader HR call, which they describe as shared by every track.
02Online assessment (track-split)
HackerRank, timed.
Quant research: 11 math, statistics and probability questions (older accounts: 15 questions in 60 min). Developer and ML engineer: 120 min, 2 hard algorithm problems
The test you get depends on the posting. The research version mixes variance of a density, Bayes, conditional expectation, limiting distributions, information gain and decision-tree splits, logic puzzles, data-sufficiency items and reading a chart or regression output; one item asks which least-squares solver suits a given system. The developer version gives two algorithm problems from a recurring bank (grid BFS and 0-1 BFS, prefix sums with binary search, amortized stacks, simulation) where brute force times out.
What it tests
- Conditional probability and expectation at roughly four minutes per question
- Information gain and choosing a decision-tree split
- Reading regression output and choosing GD, SVD or QR for least squares
- For the developer test: optimal-complexity graph and data-structure code
How to prepare
- Ask the recruiter which version you will sit; preparing for the wrong one is the most common mistake in compiled accounts.
- For the research test, bank the questions you see instantly and come back to the rest.
- For the developer test, get the complexity right before coding; timeouts are the usual failure.
Free reported questions for this stage
The firm's recruitment page says some roles include an assessment. Platform, length and content come from compiled accounts of the quant research and developer tests; no first-hand ML report covers it.
03One-way recorded video interview
HireVue, same prompts for everyone: about 15 min of think time, 6-9 segments with 1-2 min answers; scored as part of the online assessment
A motivation and communication screen rather than a technical one. Prompts include why the firm and why trading, explaining your favourite project to a non-technical person, a failure and how you recovered, and handling pressure. It rolls into your test score instead of being a separate gate.
What it tests
- Explaining a technical idea simply
- Motivation for the firm and for trading
- A failure story with a clear recovery
How to prepare
- Rehearse 60-90 second answers with a clear structure; the clock cuts you off.
- Prepare a plain-language version of your main ML project with no jargon.
No first-hand ML report mentions it. Format from compiled accounts that place the recorded video in every track, including developer roles.
04Technical round: probability, statistics and coding
45-60 min video call with a researcher: probability and statistics with derivations, plus one coding problem (often a 2-D dynamic program or combination sum)
The research track's main technical gate. Probability questions lean on conditional expectation, Markov-chain recursions and Bayesian updating, and you are expected to set up the recursion or posterior explicitly rather than quote an answer. A clean coding problem sits next to the maths, and some variants add a Monte Carlo or optimization segment.
What it tests
- Conditional expectation and first-step analysis on Markov chains
- Bayesian updating with sequential information
- Turning a probability or combinatorics setup into working code
- Monte Carlo checks and simple optimization
How to prepare
- Write the state and the recursion first; most of these collapse to one equation.
- Practise 2-D dynamic programming until the table setup is automatic.
- Offer a quick simulation to sanity-check a closed-form answer when time allows.
Format and content from compiled accounts of the quant research technical round; no first-hand ML report describes it. The problems placed here carry quant research tags at the firm.
05ML fundamentals and research depth
Inside the technical round for ML-leaning research candidates, or as its own interview with researchers; spoken questions plus an ML coding task
Where ML candidates are tested most directly. Reported questions cover information gain and decision-tree splits, random forests versus neural networks, the benefit of residual connections, classification basics, bias-variance and regularization, and when a regression beats a transformer. An ML coding task and probing on your own research follow. The firm's ML page describes large, noisy, multi-asset data and experiments that get live-market feedback within hours.
What it tests
- Tree models: information gain, splits, forests versus neural networks
- Deep learning basics such as why residual connections help
- Bias-variance, regularization and when a simple model beats a complex one
- Validation on noisy, time-ordered data: leakage, walk-forward splits, backtest pitfalls
- Defending the design and results of your own research
How to prepare
- Be ready to argue for a regularized linear baseline before reaching for a transformer.
- Explain how you would split time-ordered market data so no future information leaks into training.
- Bring outlier handling and robust statistics into data answers; the firm's tagged problems include it.
Free reported questions for this stage
- Information Gain of a Decision-Tree SplitStatistics & InferenceEasySolve
- Remove Outliers, Then Mean and StdStatistics & InferenceEasySolve
Two ML-role reports (a 2024 ML engineer, a 2022 ML researcher) describe ML and deep learning questions. Further topics from compiled accounts of the research technical round.
06Coding and systems (ML engineer and developer track)
Virtual or on-site: live coding sessions of up to 2 hours plus a systems deep dive; some Python developer roles use a 3-5 hour take-home with a README
The engineering track's onsite. Live coding rewards a fast correct solution followed by an optimization discussion, since the timeout traps from the test recur. The systems round goes deep on CPU caches and the TLB, segfaults, signals and signal masks, C++ internals, Linux debugging and REST or asyncio. Some systems onsites simulate triaging a server outage. Take-home graders look at engineering judgment, library choices and the README.
What it tests
- Optimal-complexity live coding with clean code
- Memory hierarchy, signals and C++ internals
- Debugging a Linux production issue
- For ML engineers: training and serving pipelines, GPU and distributed compute
How to prepare
- Have a 30-second explanation ready for cache misses, the TLB, signal masks and lock contention.
- Justify library choices in a take-home README; graders weigh judgment, not just a passing script.
- Know the cost of moving data between CPU and GPU and across machines.
One 2024 ML engineer report covers ML infrastructure questions. Coding and systems format from compiled accounts of the developer onsite; the problem placed here carries a software engineer tag.
07Superday / final rounds
On site (the firm arranges travel) or virtual, about five rounds: rapid probability questions at about 4 min each, about 1 hr of brain teasers, a 1 hr coding block, game theory, estimation and behavioral
The final step for research candidates, often booked within days of the technical round. Rounds mix fast probability, brain teasers, coding, game-theory puzzles such as a turn-taking delivery game, Fermi estimates with a range, and the recurring why-the-firm questions. One ML researcher instead had a case study as the final step after an interview with a senior trader.
What it tests
- Fast probability under time pressure
- Game theory by backward induction or invariants
- Structured estimation with a confidence range
- A consistent motivation story
How to prepare
- Pace the rapid-fire round: answer what you see at once, flag the rest and move on.
- Give estimates as a range with the main driver stated.
- Keep your motivation answers consistent with what you said at the HR screen.
One 2022 ML researcher report ends with a case study. The firm's recruitment FAQ mentions onsite assessment days; round mix from compiled accounts of the quant superday.
Questions and themes reported most often
- Random forests versus neural networks (1 reports, ml fundamentals and research depth)
- Why residual connections help deep networks (1 reports, ml fundamentals and research depth)
- How ML makes money in trading: results with Sharpe or P&L (hr screen)
- Why the firm and why trading, asked at every stage (hr screen)
- Information gain of a decision-tree split (ml fundamentals and research depth)
- When a regression beats a transformer on noisy data (ml fundamentals and research depth)
- Remove outliers, then report robust summary statistics (ml fundamentals and research depth)
- Choosing gradient descent, SVD or QR for least squares (online assessment (track-split))
What IMC Trading asks about
- Probability & expected value25%
- ML fundamentals & research depth25%
- Coding & algorithms20%
- Statistics & regression13%
- Systems & ML infrastructure10%
- Behavioral & motivation7%
IMC Trading interview prep checklist
- Confirmed with the recruiter which online test version applies to your role
- Solved a batch of conditional-expectation and Markov-chain problems with written recursions
- Can compute information gain by hand and compare forests, boosting and neural networks
- Can explain walk-forward validation and how to spot leakage in a feature pipeline
- Have two projects ready with method, result, baseline and a money or latency number
- Rehearsed 60-90 second recorded answers, including ML for a non-technical listener
- ML engineers: reviewed caches, the TLB, signals and why a GPU training job is slow
Tips
- Find out early which test you will sit: research-track ML roles most likely get the quant research test, ML engineer roles the developer test.
- Drill conditional expectation, Bayesian updating and Markov-chain recursions until the setup takes under a minute.
- Know tree models cold: entropy, information gain, splits, forests versus boosting versus neural networks.
- Prepare to defend a simple, regularized baseline and to explain leakage-free, time-ordered validation on market data.
- Put a number on every project you discuss: accuracy, Sharpe, P&L or latency, plus the baseline it beat.
- For ML engineer roles, add systems depth: memory hierarchy, signals, C++ internals and GPU or distributed training bottlenecks.
- Rehearse a jargon-free explanation of your ML work for the recorded video.
- Expect a fast process but volatile headcount; keep other processes moving and tell the recruiter about deadlines.
See the rest of the IMC Trading loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the IMC Trading loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
IMC Trading interview FAQ
How many interview rounds does IMC Trading have for ML roles?
Candidates report 7 stages: HR screen, Online assessment (track-split), One-way recorded video interview, Technical round: probability, statistics and coding, ML fundamentals and research depth, Coding and systems (ML engineer and developer track) and Superday / final rounds. An application (the firm asks you to pick one role and uses an AI tool to organize resumes in some locations), an HR screen of 10-45 minutes before or soon after the test, a timed online test that splits by track (quant research: 11 math, statistics and probability questions, or in older accounts 15 questions in 60 minutes; developer and ML engineer: 120 minutes for two hard algorithm problems), a one-way recorded video scored as part of that test, then a 45-60 minute technical round and a superday of about five rounds on site or virtual. Engineering candidates get two-hour live coding sessions with a systems deep dive, or for some Python roles a 3-5 hour take-home. The market-making and betting game, the mental-math and number-sequence tests and the 2026 pattern-plus-personality screen are trader-track only. With two first-hand ML reports, treat this as the quant research loop adapted for ML titles.
How long does the IMC Trading ML interview process take?
Reported timelines run Fast once it starts: compiled accounts describe superdays booked within days of the first round, though seats are often frozen or filled.
What does IMC Trading ask in ML interviews?
By reported question mix: Probability & expected value 25%; ML fundamentals & research depth 25%; Coding & algorithms 20%; Statistics & regression 13%; Systems & ML infrastructure 10%; Behavioral & motivation 7%. The most repeated themes: Random forests versus neural networks; Why residual connections help deep networks; How ML makes money in trading: results with Sharpe or P&L.
How should I prepare for the IMC Trading ML interview?
Find out early which test you will sit: research-track ML roles most likely get the quant research test, ML engineer roles the developer test. Drill conditional expectation, Bayesian updating and Markov-chain recursions until the setup takes under a minute. Know tree models cold: entropy, information gain, splits, forests versus boosting versus neural networks.
Are there free IMC Trading interview questions?
Yes. LeetML has 6 questions reported from IMC Trading interviews, and 3 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- OptiverFor ML, AI and research roles: a long timed assessment battery, a live betting game with a trader, a probability whiteboard, a P&L-graded take-home and a blunt fit final.
- Flow TradersFor ML, AI and research roles: a stats-and-ML or language-and-coding online test, a recruiter call, probability and regression interviews, a codepair for engineers, then a research deep dive with a team lead.
- Jane StreetFor ML, AI and research roles: multi-part coding judged on clarity, a hands-on ML round, and probability and statistics on research-leaning loops.
- Hudson River TradingFor ML, AI and research roles: a recycled timed coding test, a probability or programming screen, then a final day of code, systems and statistical depth on very noisy data.
- DRWFor ML, AI and research roles: a timed Python and ML test, then probability, regression and an ML design or research deep dive over a two-day final.
- Akuna CapitalFor ML, AI and research roles: a proctored Python coding test, a camera-on math test, probability interviews, then a final round on statistics, ML and your research.