Careers/Dubai, on-site/34 open
From research question to live system — you own it.
Seven disciplines, one standard. Every role has its own page: the work, the evidence we look for, and an honest account of who it would not suit. Read it, then write to us as you would to a colleague.
Ownership is the offer.
There is no research team writing specifications for an engineering team. You find the idea and try to break it. You build the system that runs it. You answer for it at three in the morning. That is true of every role on this page, and it does not suit everyone.
We publish how we work and never what we trade. Judge the standard of evidence before you apply — nothing here asks you to take our word for anything.
Quantitative Research and Strategy Development
Find the edge. Disprove it. Run what survives.
Trading and Execution
Know when the automation is wrong.
Software Engineering and Core Technology
The systems without which research cannot trade.
Data Engineering and Machine-Learning Infrastructure
The layer that sets the ceiling on every result above it.
Risk Management
Survive the event nobody modelled.
Operations, Post-Trade and Support
Make every trading day auditable.
Enabling Roles
Who we hire next — and how their time is spent.
Every position is full-time and on-site in Dubai. Compensation is discussed in the technical conversation, not published as a band we would then negotiate away from.
Four steps. We pay for the one that costs you a day.
No take-home that is really a fortnight. No panel of six. No abstract puzzles about manhole covers. The assignment is one day, drawn from our real backlog, and we pay for it whether or not we make an offer.
If a stage is going badly we say so during it rather than afterwards. You will not wait three weeks to learn nothing.
- IntroductionA short message and something you built. No cover letter, no recruiter, no form.
- Technical conversationNinety minutes on work you actually did, deep enough that neither of us can bluff.
- Paid assignmentA real problem from our backlog, scoped to one day. We pay for it either way.
- ReviewYou defend the work to the people who would maintain it. This is the decision.
A researcher tests a thousand configurations over the same five years and reports the best one. It shows an annualised Sharpe ratio of 2.1.
Assume every configuration has zero true edge. What is the distribution of the largest Sharpe ratio across those thousand trials? At what number would you start to believe it? And what would you rather know about the search than the number itself?
Send the reasoning, not the answer.
puzzle (at) mansastrategies.com
Register for puzzle updates.
Optional. Leave an email, LinkedIn URL, or X profile. We store a device fingerprint only when you tick the box and submit — never on page load, never silently. Used solely to recognise people interested in the puzzle race.