Company interview guide
Old Mission Capital Machine Learning Interview Guide (2026)
For ML, AI and research roles: a 25-minute, 35-question probability speed test, a behavioral round, then probability, statistics and coding rounds at a multi-asset market maker.
Old Mission Capital's machine learning interview loop has 7 reported stages, from application and track choice to final rounds; typical length: Several weeks; timeline not reported first-hand.
Roles covered: Research Scientist, Applied Scientist, ML Engineer and Data Scientist. The question bank holds 3 questions reported from Old Mission Capital interviews, 1 of them free to solve in your browser.
- Interview process length
- Several weeks; timeline not reported first-hand
- Stages
- 7
- Reports linked to bank questions
- 3
- Bank report dates
- Jun 2024–Jul 2026
- Questions in bank
- 3 (1 free)
- Last reviewed
- October 2026
How the Old Mission Capital interview loop works
Apply through the firm's job board, then a timed online assessment: a fixed 35-question multiple-choice probability test in 25 minutes on the research and trading track, or a Codility set of about two hours (CS multiple choice plus about two coding problems) on the engineering track. Passing leads to a behavioral round, then technical rounds of about 45-60 minutes each on probability, expectation and mental math, with statistics and research discussion for researchers, and final rounds that push the hardest probability and sizing problems; engineers get C++ and systems instead. The options and market-making round, quote-a-market exercises and game-EV brain teasers are trader-track only and get no stage here. The outline rests on compiled accounts, not first-hand ML reports.
The Old Mission Capital interview process, stage by stage
01Application and track choice
Online application on the firm's job board; research roles in Chicago and New York, engineering roles in Chicago, New York and London
There is no ML or AI title. ML researchers apply as quantitative researchers: the systematic equities researcher sits on an ML team in New York (4+ years; alpha signals on structured and unstructured data, feature engineering, model validation, backtesting and out-of-sample testing), while the options researcher roles in Chicago, including a Ph.D. graduate program, build and calibrate pricing models for volatility products in Python and C++. ML engineers fit the research infrastructure engineer role (data pipelines, firm-wide data capture and indexing, cluster tooling, moving data between C++ and Python) or the software engineer graduate and intern programs.
What it tests
- A quantitative degree: math, statistics, physics, computer science or engineering
- For the ML team: a record of finding and validating alpha signals in equities
- For options research: derivatives and volatility-surface modeling, Python and C++
- For research infrastructure: C++, Python and big-data pipelines
How to prepare
- Pick the posting by what you will be tested on: research roles lead to the probability test, engineering roles to the coding set.
- Lead the resume with validation results (out-of-sample, after costs) rather than model architectures.
- The firm recruits on campus in the fall (2026 visits included several large engineering schools); meeting the team there is the route its careers page suggests.
The firm's job board and research postings describe the roles and required background; no posting describes the interview stages and no first-hand ML-role report covers this step.
02Online assessment
Research and trading track: fixed 35-question multiple choice in 25 minutes (about 43 seconds each), same questions in the same order for everyone, sometimes with a short coding task; engineering track: Codility, about 2 hours, CS multiple choice plus about 2 coding problems
The gate before any interview. The math version is a speed test of recognition: dice sums, handshake counts, expected profit with defects, two jars of the same colour, Bayes posteriors (a double-headed coin, an unsafe driver after an accident), pattern races such as HH before TH or HHH before TTTT, and Markov-chain tails like a candy conveyor belt. Some math sittings add a short coding problem, such as a recursive-descent parser for nested lists of strings. The engineering set mixes CS questions with digit, string, bitmask and prefix-sum problems.
What it tests
- Counting and probability templates solved in under a minute
- Bayes updates with a prior and one or two observations
- First-step recursion for pattern races and stopping times
- Stationary distributions of small Markov chains
- Engineering track: correct, efficient code under an autograder
How to prepare
- Pace at about 43 seconds a question; skip anything that needs a second line of algebra and come back.
- Use the answer options: bound the expected value first and eliminate impossible choices.
- For pattern races, track which suffix of each target you currently hold and write one equation per state.
- Ask the recruiter which version you will sit; research and engineering candidates have reported different tests.
Format, timing and question themes from compiled accounts of the firm's assessments, consistent over 2021-2026; the one placed problem is tagged to an engineering report. No ML-role report.
03Behavioral round
Conversation after the assessment; length and interviewer not reported
Compiled accounts place a behavioral round between the assessment and the technical rounds. The firm's careers page stresses enjoying puzzles and new technology, competitive drive, early responsibility, collaboration and intellectual humility, so expect questions on why markets, why a mid-sized firm, and how you work with traders and engineers.
What it tests
- Why trading research rather than tech or academia
- Collaboration with traders and engineers
- Ownership of a project end to end
- Handling being wrong under time pressure
How to prepare
- Have a two-minute summary of one research or engineering project with a measurable result.
- Give a concrete answer on why markets: fast, honest feedback on models that go live.
- Ask what the technical rounds cover; the answer tells you which track you are on.
Only compiled accounts of the firm's process mention this round after the assessment; the careers page describes what the firm looks for but not the round itself.
04Probability and mental math round
45-60 minutes by video with follow-ups
The core technical round for researchers and traders alike. Reported themes are dice and card draws, conditional events, gambler's ruin and biased random walks, expected flips until consecutive heads, rolling until a six, last-ball and first-ace problems, fair entry fees for simple games, and Poisson arrivals such as quote hits across venues. Interviewers extend each answer: once you have the probability, expect to be asked what the bet is worth and how the answer changes with one parameter.
What it tests
- Conditioning on the first step and solving for a fixed point
- Linearity of expectation and symmetry before casework
- Random walks and ruin probabilities
- Poisson processes, thinning and exponential waiting times
- Fast, accurate arithmetic on fractions and expected values
How to prepare
- Estimate the answer aloud, then formalize; interviewers grade the route as well as the number.
- Write the recursion explicitly (E = 1 + p E and its relatives) instead of guessing.
- Sanity-check against small cases and limits before you state a result.
- Drill mental arithmetic daily; this round is timed even when it feels conversational.
Format and themes from compiled accounts of the firm's quant research and trading loop; no first-hand ML-role report, and no visible problem tagged to the firm fits this round.
05Statistics, Bayesian inference and research discussion (research track)
Part of the technical and final rounds for researchers; separate length not reported
The research track adds statistics to the probability rounds. Reported items include a Beta posterior for a biased die, a confidence interval and p-value for a poll, a t-test for whether an ETF premium persists, tracking-error tail probability and VaR, posterior fair value from a noisy signal, sequential Bayesian updating on trade direction, a joint-normal conditional probability, and a regime-switching liquidity chain. For the equities ML team, the posting adds feature engineering, model validation, backtesting and out-of-sample testing, so expect to defend a signal's robustness.
What it tests
- Conjugate Bayesian updates and posterior tail probabilities
- Confidence intervals and tests on proportions and means
- Conditional expectation with jointly normal variables
- Whether a signal survives out-of-sample and after costs
- Bet sizing from an edge estimate (Kelly and fractional Kelly)
How to prepare
- Know the Beta-binomial and normal-normal updates cold, including posterior mean as a shrunk average.
- When testing persistence in a time series, say how autocorrelation changes the standard error.
- For a sizing question, derive full Kelly, then explain why you would trade half of it given estimation error.
- Describe validation in time order: walk-forward splits, no look-ahead in features, and a count of variants tried.
Free reported questions for this stage
- Bayesian Posterior for a Biased DieStatistics & InferenceEasySolve
Themes from compiled accounts of the firm's quant research questions and the ML-team posting's duties; both placed problems carry research-track tags. No first-hand ML-role report.
06Coding and systems rounds (engineering track)
Live coding and systems interviews for engineers, C++-focused; number and length not reported
ML engineers applying as software or research infrastructure engineers face C++ and systems rounds in place of the trader games. Compiled question lists add numerical coding that suits ML engineers: online mean, variance and covariance in O(1) per update, a trailing time-weighted average price in O(n), counting subarrays whose average return clears a threshold, and maximum profit from a premium series. The research infrastructure posting points to data pipelines, storage and indexing at firm scale, and C++ to Python data movement.
What it tests
- Streaming statistics with constant memory and numerical stability
- Sliding-window and prefix-sum techniques on time series
- C++ fundamentals: memory, performance and clean interfaces
- Designing a data pipeline that captures, stores and indexes market data
How to prepare
- Implement Welford-style updates and explain why the naive sum-of-squares formula loses precision.
- Turn average-above-threshold problems into prefix sums with a shifted target.
- Be ready to talk through zero-copy hand-off between C++ and Python and columnar storage choices.
Engineering-track content from compiled accounts of the firm's loop plus the research infrastructure posting's duties; no first-hand report from an ML engineer.
07Final rounds
Several back-to-back interviews; on site or virtual not reported for research roles
The closing rounds draw the hardest reported problems: multi-step probability, pattern races and Markov tails at full depth, sizing under constraints such as a drawdown limit, adverse selection in a spread, and coding. Researchers may be asked to discuss past work or a modeling choice in depth. Trader candidates see game-EV teasers and market-making mental math here, which research candidates usually do not.
What it tests
- Hard probability with long follow-up chains
- Optimal sizing under constraints
- Spread setting with informed and noise traders
- Defending a past research project's validation
How to prepare
- Keep probability practice going right up to the finals; the hardest items appear here.
- When a problem turns into a market, state the price you would quote and the edge it leaves.
- Prepare one project you can defend to a skeptical researcher for 20 minutes.
Compiled accounts of the firm's final rounds describe the hardest probability, sizing and coding problems; research-track format and length are not reported first-hand.
Questions and themes reported most often
- Bayes posterior after an observation: a double-headed coin, an unsafe driver after an accident (online assessment)
- Pattern races by first-step recursion: HH before TH, HHH before TTTT (online assessment)
- Beta prior on a die or coin: posterior, tail probability and MAP (statistics, bayesian inference and research discussion (research track))
- Intervals and tests: poll proportion, persistence of a premium (statistics, bayesian inference and research discussion (research track))
- Fair value from a noisy signal with jointly normal variables (statistics, bayesian inference and research discussion (research track))
- Gambler's ruin and biased random walk hitting probabilities (probability and mental math round)
- Kelly and fractional Kelly sizing for a signal, with a drawdown limit (statistics, bayesian inference and research discussion (research track))
- Online mean, variance and covariance in constant time per update (coding and systems rounds (engineering track))
What Old Mission Capital asks about
- Probability & expectation35%
- Statistics & Bayesian inference25%
- Signal research & validation discussion15%
- Coding (OA and engineering track)15%
- Mental math & speed10%
Old Mission Capital interview prep checklist
- Scored 30 or more of 35 on a timed 25-minute probability multiple-choice set
- Can solve HH before TH and HHH before TTTT with state equations in under two minutes
- Can derive a Beta-binomial posterior and its tail probability without notes
- Can build a confidence interval and p-value for a proportion and explain autocorrelation's effect
- Can derive full Kelly for a simple edge and argue for a fractional bet
- Can explain walk-forward validation, look-ahead leakage and a multiple-testing correction for a backtest
- Engineers: implemented online mean, variance and covariance with stable updates
- Have a clear answer to why a market maker over an ML role in tech
Tips
- Find out early which assessment you will sit; the math test and the engineering set need different preparation.
- Drill the 35-in-25 format against a clock: recognition of the template beats derivation.
- Write Bayes updates in odds form and pattern races as one equation per state.
- Know conjugate updates, interval construction and the effect of autocorrelation on standard errors.
- Tie closed-form answers to ML practice: shrinkage, regression on noisy signals, sizing against an overfit edge.
- Describe research validation in time order with a count of variants tried.
- Skip the options quoting and betting games unless your recruiter says your track includes them.
- Engineers: practise streaming statistics and C++ to Python data hand-off, not only algorithms.
See the rest of the Old Mission Capital loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Old Mission Capital loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Old Mission Capital interview FAQ
How many interview rounds does Old Mission Capital have for ML roles?
Candidates report 7 stages: Application and track choice, Online assessment, Behavioral round, Probability and mental math round, Statistics, Bayesian inference and research discussion (research track), Coding and systems rounds (engineering track) and Final rounds. Apply through the firm's job board, then a timed online assessment: a fixed 35-question multiple-choice probability test in 25 minutes on the research and trading track, or a Codility set of about two hours (CS multiple choice plus about two coding problems) on the engineering track. Passing leads to a behavioral round, then technical rounds of about 45-60 minutes each on probability, expectation and mental math, with statistics and research discussion for researchers, and final rounds that push the hardest probability and sizing problems; engineers get C++ and systems instead. The options and market-making round, quote-a-market exercises and game-EV brain teasers are trader-track only and get no stage here. The outline rests on compiled accounts, not first-hand ML reports.
How long does the Old Mission Capital ML interview process take?
Reported timelines run Several weeks; timeline not reported first-hand.
What does Old Mission Capital ask in ML interviews?
By reported question mix: Probability & expectation 35%; Statistics & Bayesian inference 25%; Signal research & validation discussion 15%; Coding (OA and engineering track) 15%; Mental math & speed 10%. The most repeated themes: Bayes posterior after an observation: a double-headed coin, an unsafe driver after an accident; Pattern races by first-step recursion: HH before TH, HHH before TTTT; Beta prior on a die or coin: posterior, tail probability and MAP.
How should I prepare for the Old Mission Capital ML interview?
Find out early which assessment you will sit; the math test and the engineering set need different preparation. Drill the 35-in-25 format against a clock: recognition of the template beats derivation. Write Bayes updates in odds form and pattern races as one equation per state.
Are there free Old Mission Capital interview questions?
Yes. LeetML has 3 questions reported from Old Mission Capital interviews, and 1 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.
- Susquehanna (SIG)For ML, AI and research roles: a timed probability test, regression-heavy technical calls, a trade-data take-home and a long onsite.
- 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.
- IMC TradingFor 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.
- Five RingsFor ML, AI and research roles: a camera-on no-calculator math screen, two live probability and statistics rounds, then a research-track data project.
- 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.
Official sources
- Firm site: what we do (official)
- Firm site: working at the firm (official)
- Firm site: open positions (official)
- Firm posting: Quantitative Researcher, Systematic Equities (official)
- Firm posting: Quantitative Researcher (Ph.D.) graduate program (official)
- Firm posting: C++ Software Engineer, Research Infrastructure (official)