From raw ticks to risk-managed fills.
Data
Tick-level market data across venues, cleaned, aligned and stored once, then used for research and live inference alike. One pipeline, no train/serve skew.
Models
Deep and probabilistic models trained on order flow and microstructure. We forecast distributions rather than point estimates, and size every position against its own uncertainty.
Execution
Signals become orders through execution logic that models its own market impact, fill probability and queue position, tested against tick-by-tick replays of history.
Risk
Exposure limits, drawdown circuit-breakers and kill switches live outside the models. They run continuously and cannot be overridden by the strategies they govern.
Built like infrastructure, not a notebook.
Research-grade models, production-grade engineering. The pieces behind every decision.
Probabilistic modelling
We forecast the full distribution of outcomes, not a single number. Position size follows conviction and uncertainty, so the system takes risk where the edge is and steps back where it isn't.
Research rigour
Point-in-time data and strict out-of-sample validation before anything touches capital.
One data pipeline
Full-depth market data normalised once, shared by research, simulation and production.
Execution intelligence
Order placement that models its own impact, queue position and fill probability, and replays history tick by tick.
Independent risk layer
Limits, circuit-breakers and kill switches sit outside the models. Every decision is logged, attributable and replayable.
Researchers and engineers from the top of the field.
Our team comes from leading Wall Street hedge funds and from the research groups of MIT, UC Berkeley and peer institutions. We pair institutional market experience with frontier machine-learning research.
Let's build what comes next.
We're always interested in talking with exceptional researchers, engineers and partners.