// approach
We don't forecast headlines. We build a repeatable pipeline that converts raw market data into measured, risk-budgeted positions - and we improve that pipeline every single day.
// the pipeline
Data → Signal → Risk → Execution
Every dot is a hypothesis entering the pipeline. Watch where they fall - most ideas die, and that's the system working. Hover a stage to trace its gate.
Clean, versioned, point-in-time-correct market and alternative data. Most quant failures are data failures in disguise, so this layer gets obsessive attention.
✝ dies here: look-ahead bias, survivorship, vendor errors
Every hypothesis fights for its life: out-of-sample validation, regime slicing, decay analysis. Most ideas die here - and that's the system working.
✝ dies here: fails out-of-sample - most ideas were noise
Hard drawdown budgets, factor-aware construction, convex tail protection. The portfolio is engineered to survive the day we're wrong.
✝ dies here: too correlated, too crowded, tail too fat
Microstructure-aware execution that measures every basis point of cost. Alpha is discovered in research - it is kept or lost in execution.
✝ dies here: costs eat the edge at realistic size
// operating principles
The rules we never trade against
$ Evidence over ego
If the data disagrees with the story, the story loses.
$ Risk is a budget
Spent deliberately, never discovered after the fact.
$ Small edges, compounded
We industrialise many modest edges instead of hunting one miracle.
$ Automation with judgment
Machines execute; humans own the assumptions.
$ Capacity honesty
Every strategy has limits - we measure and respect them.
$ Kill switches everywhere
Anything live can be halted instantly, by anyone on the desk.