Teza Technologies Interview: Official Process and Role Guide

Start with the Teza that exists now

Teza was founded in 2009 by Misha Malyshev after he left Citadel's high-frequency business. It now describes itself as a systematic trading firm translating mathematical ideas into strategies on major markets, with holding periods ranging from milliseconds to months. That is broader than the familiar "Chicago HFT" label. A candidate who prepares only for nanoseconds and lock-free queues may miss the research horizon, product and team named in the vacancy.

The firm's current locations page lists Austin, New York, Chicago, London and Yerevan. Chicago houses technology, legal and compliance, and operations. Investment teams are mentioned in Austin, New York and London; London also houses risk and operations. Current openings are concentrated in Austin, London and Yerevan, although that mix changes.

Teza Technologies and its affiliate Teza Capital Management are not interchangeable measures of scale. Regulatory AUM for an investment adviser is not the gross capital or trading activity of the wider technology and trading group, so this guide does not attach a speculative firm-wide AUM or headcount to Teza.

What Teza publishes about its interviews

Teza's own candidate FAQ is the useful source here. It says:

• technical ability may be assessed through coding tests, live problem-solving and technical discussions;
• non-technical skills are assessed in live interviews;
• relevant topics may include linear algebra, algorithms, machine-learning basics, architecture, software design, the software-development lifecycle and structured reasoning;
• recruiters tell candidates which topics to expect in advance; and
• the average process lasts 25 to 30 days, depending on the role.

The careers page also says Teza uses an automated employment decision tool to assess role-relevant technical skills. It expressly asks candidates not to use AI or external tools during assessments.

That is the official common denominator. It does not lock every candidate into two phone screens, a one-to-two-week research take-home, six Chicago interviews and an offer on a set clock. Candidate reports may describe some of those elements. They vary by office, seniority and team. Treat your recruiter's briefing as the specification.

Read the vacancy through Teza's formula

Teza publishes an idiosyncratic hiring formula on its Who We Are page: IQ, IT and HR are threshold requirements, while "Drive" is exponential. Behind the branding is a practical interview rubric.

IQ means structured mathematical reasoning. State assumptions, reduce the problem and test edge cases. A fast answer with an unspoken error is weaker than a compact derivation that can be checked.

IT means hands-on technical competence. Teza says most employees write code, but the expected level changes by role. Its public list includes C++, Java, Python, Excel automation and integration work; it does not nominate one universal house language.

HR means trustworthy communication. Explain your work without hiding uncertainty. Listen when an interviewer changes a condition. The firm explicitly includes tone, clarity and responsiveness.

Drive means ownership. Choose examples in which you carried a difficult piece of work through to a usable result, including the unglamorous validation and operational details.

Choose a stream before choosing revision topics

Teza's stated range from milliseconds to months makes role selection unusually important. Begin with the vacancy's stream, instruments, horizon and office. Then write down the main decision the role improves. A short horizon researcher may be improving a forecast or execution decision from event data. An options researcher may be improving valuation, volatility or hedging. A platform engineer may be improving the speed and reliability with which many teams can test or trade ideas.

That decision should govern your preparation. Do not treat probability, machine learning and systems design as interchangeable badges. If the work concerns short horizon data, be ready to discuss event ordering, dependence, fills and the economic meaning of a tiny edge. For a slower systematic strategy, focus on feature availability, portfolio exposures, turnover, capacity and regime stability. For engineering, identify whether the user is a researcher, an automated strategy or an operational team, then explain the consequence of stale, late or incorrect output.

Candidate fit is clearer when evidence mirrors the role. Select one difficult problem where you made the central judgement, one implementation you can inspect in detail and one occasion when contrary evidence changed your plan. State which parts were yours and which belonged to colleagues. Teza's formula gives you a useful test: the example should contain reasoning, technical execution, communication and sustained ownership rather than merely naming all four qualities.

Quant research: demonstrate an honest experiment

Teza currently advertises quantitative research across several streams and locations, including an all-streams role in Austin and options research in London. The homepage emphasises research, rapid prototyping and converting discoveries into traded strategies.

The best preparation artefact is therefore one experiment you know intimately. Be able to reconstruct when every feature became available, how hyperparameters were chosen, which costs were included and what happened in an unfriendly regime. If the work is high frequency, discuss event-time sampling, queue position and timestamp integrity. If it is slower, do not bolt on low-latency jargon that the strategy does not need.

Useful earlier published revision includes the quant research interview guide for experimental reasoning and the market microstructure guide for order driven markets. Choose the relevant sections according to the role rather than treating either as a rote reading list.

Show a research standard, not just a result

A Teza research answer should make it possible for another person to decide whether the result deserves more work. Start with a falsifiable economic or market hypothesis. Describe the point in time dataset, labels and realistic information set. Separate choices made before seeing results from those made after. Report failed variants and the number of ideas explored, since a best backtest without its search history exaggerates the evidence.

Then connect the forecast to trading. Explain how scores become positions, how correlated bets are controlled and whether performance survives plausible delay, spread, fees and impact. If the strategy trades quickly, the fill model and timestamp chain may dominate the model improvement. If it holds for months, unstable exposures and crowding may matter more. In both cases, show sensitivity ranges rather than one attractive cost assumption.

Finally, define a next decision. You might reject the idea, collect a cleaner dataset, run a shadow test or allocate tightly limited risk. Saying what evidence would reverse that choice demonstrates structured reasoning. It also avoids presenting research as a sequence in which every hypothesis conveniently succeeds.

Engineering and infrastructure: locate the constraint

Teza separates engineering, infrastructure and real-time-platform openings on its careers page. That distinction should shape your examples.

For trading systems, you may need to reason about throughput, tail latency, memory ownership and failure recovery. For research platforms, reproducibility, data lineage and safe deployment may matter more. Infrastructure interviews can range from Linux and networks to capacity planning and incident response. Ask the recruiter which layer the team owns before deciding that FPGA, kernel bypass or any particular vendor technology is compulsory.

An effective systems answer begins with the service-level objective and failure model. Only then select concurrency, storage and deployment patterns. Teza's published topics include architecture, software design and SDLC, so testing and release judgement deserve as much attention as clever code.

Original practice problems

The following are practice exercises, not claimed Teza interview questions.

Research: too many observations

A one-minute forecast has a t-statistic of 4.5 over a year. What would make you distrust it?

Discuss serial dependence, overlapping labels, the number of strategies tried, revisions to the universe and costs. Estimate an effective sample size rather than treating every minute as independent. Then specify the next falsification test and what result would make you stop.

Research: execution changes the answer

A backtest buys at the mid-price one second after each signal. How would you rebuild it?

Use observable quotes, latency assumptions, spread, queue or fill logic, fees and market impact appropriate to the product. Separate a forecast test from an executable strategy test. Sensitivity to latency and participation is more revealing than a single "conservative" cost deduction.

Software: replay without ambiguity

Design an order-book replay component that receives add, cancel and execute messages and must recover after a restart.

Clarify sequence guarantees and exchange semantics. Explain the order-ID index, price-level structure, snapshots, gap detection, idempotence and deterministic replay tests. A monotonic happy-path implementation is only the beginning.

Architecture: research to production

How would you promote a new model while ensuring that researchers can reproduce the decision made by yesterday's live system?

Cover immutable data and feature versions, model artefacts, configuration, approvals, shadow runs, rollback and decision logs. Tie every control to a plausible failure rather than reciting platform components.

You can practise implementation under time pressure with the Quantt coding tests, but use Teza's advance topic list to choose relevant exercises.

Questions worth asking the recruiter

Teza expressly encourages candidates to use the recruiter conversation to clarify the role. Make that call count:

• Which team and office owns the role?
• Which technical topics and language will be assessed?
• Is the coding environment supplied, and what tools are prohibited?
• Does the work sit closer to research, real-time trading or shared infrastructure?
• How are the remaining stages likely to fit within the stated 25 to 30 day average?

These questions are more useful than trying to infer a private stack from old candidate accounts.

Frequently Asked Questions

How long does a Teza interview process take?

Teza says its average hiring process takes 25 to 30 days, depending on the role. That is an average rather than a guaranteed deadline, and extra conversations may be suggested if another position looks more suitable.

Does Teza use coding tests?

It may. The firm says technical assessment can include coding tests, live problem-solving and technical discussions. The precise mix and topics are role-dependent, and recruiters share preparation topics in advance.

Is Teza only a high-frequency trading firm?

No. Teza says its systematic strategies hold positions for periods ranging from milliseconds to months. Some current roles are explicitly HFT-focused, but candidates should follow the advertised stream rather than prepare for a single horizon.

Where does Teza have offices?

Its current official locations are Austin, New York, Chicago, London and Yerevan. Investment teams are mentioned in Austin, New York and London; Chicago houses technology, legal and operations. For how Teza sits among other systematic shops, see the proprietary trading firms guide.