Kepos Capital Interview Guide: Evidence-Based Preparation

What Kepos Actually Discloses

Kepos Capital's current official site gives four useful facts. The firm was founded in 2010, is based at 11 Times Square in New York, manages more than $2 billion and focuses on systematic macro, carbon trading and equity events.

That last point corrects a common oversimplification. Kepos is not adequately described as a factor fund trading rates, currencies, commodities and equity indices. Its own description names three strategy areas, including carbon and event-driven equity work. A candidate should identify which one the vacancy serves before deciding that PCA, Black-Litterman or the Carhart factor model must dominate the interview.

Contemporary reporting is clearer about the origin. In 2010, Institutional Investor described Mark Carhart as launching Kepos, with Giorgio De Santis as director of research and Robert Litterman chairing an academic advisory board. Neat two-founder retellings of that history are not supported by the contemporary account.

Most importantly, Kepos publishes no careers portal, standard interview stages, assessment format or question bank. We found no sufficiently detailed current candidate reports to establish one. A useful guide has to say that plainly.

A Process You Must Confirm

Some third-party guides describe a transcript screen, a 45-to-60-minute researcher call, a one-to-two-week take-home, four or five Midtown interviews, a founder meeting and a two-to-three-week decision. None of those timings or stages appears in Kepos's public material.

The official site provides a general contact address, not a recruiting workflow. A March 2026 New York portfolio-manager advert offers one current piece of role evidence: systematic, model-driven commodity and macro portfolios, with emphasis on energy and carbon commodities, plus collaboration with portfolio managers, researchers and analysts. It does not tell us how a research, engineering or graduate candidate will be assessed.

If invited, resolve the process before preparing:

• Which strategy and team owns the role?
• How many conversations are planned, and with which functions?
• Is there live coding, a written exercise or a take-home?
• What data, libraries and external sources are permitted?
• Is the exercise judged as research, production code or both?
• What is the expected time commitment and submission format?

This is not administrative fuss. At a boutique manager, a title such as quantitative researcher can cover strategy research, portfolio implementation, data work or tooling. The answer changes the preparation.

Anonymous reports would be candidate-reported, not official, even if detailed reports existed. The absence of a large public sample is itself a reason to avoid false confidence, not an invitation to fill the gap with standard hedge-fund stages.

Preparation by Strategy

Systematic macro

Build one research discussion around a cross-asset hypothesis. Define the economic mechanism, tradable instruments, forecast horizon, position normalisation and cost model. Then state where the backtest is most vulnerable: revised macro data, overlapping observations, unstable contract rolls, funding, crowding or regime dependence.

Factor knowledge is useful, but do not force equity terminology onto macro portfolios. Value might mean a real-yield or purchasing-power measure; carry depends on the instrument; momentum can be implemented across different horizons. Our arbitrage pricing theory guide is helpful for factor language, while our quant research questions covers validation.

Carbon and energy

Carbon markets are named on Kepos's home page and in the 2026 portfolio-manager advert. Prepare the institutional detail. Know which allowance or offset instrument your example uses, how compliance demand is created, what determines supply, how vintages and registries work and where liquidity sits.

A carbon backtest can fail before the model begins. Policy changes may alter the market, contracts may not be comparable through time and apparently clean price histories may hide changes in eligibility. Explain those breaks. Do not present a carbon price as one globally fungible series.

For energy-linked work, connect the physical market to the derivative: storage, transport, seasonality, basis and curve shape. A high correlation in returns is not an economic explanation.

Equity events

Kepos officially names equity events as a core expertise. Prepare a point-in-time event study with a defensible control group. Define the event timestamp, remove information that arrived after it, model overlapping events and explain how you would distinguish announcement drift from market, sector and factor exposure.

Carhart's 1997 four-factor paper is relevant background because momentum can contaminate apparent event alpha. Our Fama-French model guide covers the benchmark family. The correct benchmark still depends on the event and holding period.

Engineering and research infrastructure

There is no public evidence that Kepos uses less C++ than platform funds or a particular distributed backtesting design. Prepare from the vacancy. For research infrastructure, be ready to discuss point-in-time data, deterministic reruns, model versioning, portfolio constraints and the hand-off from research to live decisions.

One strong example is better than a guessed stack. Explain a system you built, the failure it had to tolerate, the measurement that shaped the design and what you would change now.

An Evidence-Aware Application

Kepos's limited public recruiting information makes tailoring more important, not less. Begin with the named strategy and the work in the vacancy. Build a short evidence map linking each major responsibility to a project, decision or system you can discuss. Mark the gaps honestly. Carbon-market interest supported by policy research is different from experience trading allowances; a macro backtest is different from running portfolio risk; an academic event study is different from production research.

For each research example, preserve what was known at the time. Record data releases, revisions, universe rules, model-selection decisions, trading costs and tests that failed. For engineering examples, state personal ownership, users, scale, recovery requirements and a measured result. A candidate who can reproduce a modest result and explain its limitations gives stronger evidence than one who presents a spectacular chart with no provenance.

The application should also respect confidentiality. Do not submit a former employer's data, positions or implementation. Explain the economic question, validation design and your decisions at a level that demonstrates skill without transferring intellectual property. Where a figure cannot be verified or disclosed, say that plainly.

What Interviewers Could Reasonably Evaluate

Kepos publishes no interview rubric, so these are inferences from its disclosed strategies and the collaborative language in the 2026 portfolio-manager advert, not claims about private scoring.

• Economic mechanism: can you explain why a systematic relationship might persist rather than relying on its historical fit?
• Data provenance: do timestamps, revisions, eligibility rules and contract definitions match the proposed decision?
• Cross-asset judgement: can you compare signals without assuming that value, carry or momentum means the same thing in every market?
• Portfolio awareness: do you understand covariance uncertainty, concentration, liquidity and the difference between signal strength and position size?
• Implementation discipline: can research be rerun, monitored and stopped when the evidence changes?
• Collaborative clarity: can specialists in research, portfolio management and engineering challenge the work without losing its assumptions?

Demonstrate those qualities while solving rather than reciting them. Start with a baseline, identify the decision variable and choose the test most likely to disprove the idea. If the prompt concerns carbon or energy, establish the institutional and contract details before fitting a model. If it concerns macro data, separate first releases from revised history. If it concerns events, lock the timestamp and benchmark before inspecting post-event returns.

Deciding Whether the Role Fits

A smaller systematic manager can offer close contact between research, portfolio decisions and implementation, but the exact breadth is unknown until Kepos describes the vacancy. Ask how the role divides time among original research, portfolio construction, data work and production support. Ask who can approve model changes, how strategies are reviewed and whether the seat specialises in macro, carbon or equity events.

The trade-offs differ by preference. Carbon work may provide distinctive policy and market-structure problems but requires comfort with changing rules and fragmented instruments. Systematic macro offers cross-asset breadth but exposes models to regime shifts and revised economic data. Equity events can provide cleaner hypotheses yet demand meticulous timestamp and benchmark control. An infrastructure role may touch all three while offering less ownership of final positions. Fit follows from that mandate, not from the boutique label.

Practice Prompts

These are Quantt exercises inferred from Kepos's disclosed strategies and the 2026 portfolio-manager advert. They are not reported Kepos questions.

Systematic macro: A trend signal works across rates and commodities but most P&L comes from two crisis periods. How would you test whether it is diversified, a disguised duration trade or dependent on volatility scaling?

Carbon: An emissions contract rises after a policy announcement, but open interest and eligible participants also change. Design a test that separates a genuine forecast from a market-structure break.

Equity events: Post-announcement returns appear positive. How would you choose the timestamp, benchmark and holding period without selecting them after seeing the result?

Portfolio construction: Two signals have similar standalone Sharpe ratios but correlated losses. How would you allocate risk when the covariance estimate is least reliable in stress?

Research engineering: A vendor revises economic history. Design storage and interfaces that allow researchers to use both the latest corrected series and the release available on each historical date.

Research judgement: Name the result in your own work that is most likely to disappear. What exact test would make you stop using it?

The preparation goal is not to reproduce eight polished answers. It is to show that economic mechanism, data provenance and portfolio effect remain connected.

Frequently Asked Questions

Does Kepos publish a take-home process?

No. There is no official careers page or public sequence. Any exercise, deadline or interview count should be confirmed directly for the vacancy.

Is a PhD required?

Kepos publishes no firm-wide degree rule. A quantitative postgraduate background may be relevant to some research seats, but the role advert is the only safe guide. A PhD or strong master's is not stated as a minimum in public material.

What does Kepos trade?

Its official site names systematic macro, carbon trading and equity events. A 2026 portfolio-manager advert adds model-driven commodities and macro work with an energy and carbon emphasis. It is unsafe to infer the complete portfolio or asset allocation from those disclosures.

How much does Kepos manage?

The firm currently says more than $2 billion. We use that official wording rather than larger third-party estimates with unclear dates or definitions.

Does Kepos recruit graduates?

There is no public annual graduate programme on the current site. That does not prove the firm never hires graduates; it means applicants should rely on a specific vacancy or direct contact rather than an assumed cycle.

What does Kepos pay?

There is no public firm-wide scale. The March 2026 portfolio-manager advert listed a $196,000 to $198,000 salary range for that specific New York role. It should not be extrapolated to researchers, engineers, bonuses or other years.

Kepos offers less public interview information than the larger firms. Good preparation starts by respecting that boundary, then going deeper on the strategy the role actually serves.