Research that survives contact with the market.
Backtests are cheap. We build strategies for the only environment that matters — live capital, real spreads, changing regimes — and hold every idea to the standard of surviving there. From first hypothesis to production monitoring, one team owns the entire loop.
Every strategy runs the same gauntlet.
Four stages, no shortcuts. An idea graduates only when it has earned the next stage — and most don't.
Signal discovery
Systematic search across market microstructure, cross-asset relationships and alternative data. Hypotheses are written down before testing begins, so results can't be quietly cherry-picked afterwards.
Validation
Walk-forward, out-of-sample and regime-aware testing pipelines. We treat a good backtest as a claim to be attacked, not a result to be celebrated.
Execution & risk engineering
Position sizing, smart execution and hard risk limits wired directly into the trading stack — costed at realistic spreads and slippage before capital ever touches it.
Live monitoring
Every deployed model is watched 24/7 with drift detection, alerting and automatic circuit-breakers. When live behaviour diverges from research, the system de-risks first and asks questions after.
Cross-asset, by design.
One research framework, four asset classes. The same discipline of validation and monitoring applies wherever the signal comes from.
Shipped, not simulated.
We measure the practice by what reaches production and stays there — not by what looks good on a slide.
We kill overfit ideas before they cost money.
Most of what we research never ships — deliberately. The gap between 25+ strategies engineered and 5 running in production is not attrition; it is the standard doing its job. A strategy earns capital by surviving out-of-sample validation, realistic cost modelling and live monitoring — and it keeps that capital only for as long as its live behaviour matches its research profile.
- Out-of-sample or it doesn't count. In-sample fit is a starting point, never evidence.
- Costs modelled up front. Spread, slippage and capacity are part of the backtest, not a footnote.
- Retired without sentiment. When a live strategy drifts from its research profile, it comes down before it becomes a story.
- No performance claims. Returns depend on mandate, market and sizing — we publish what we ship, not what we simulate.
Put your next strategy through the loop.
Bring us a hypothesis, a dataset or a mandate. We'll tell you — honestly — whether there's an edge worth engineering, and what it takes to run it live.