Desktop required

This experience is optimized for larger screens. Please visit on a laptop or desktop.

AI research lab automating investing.

From algo trading to reasoning-driven investing.

We are scaling autonomous research.
Markets are our verification loop:
agents turn hypotheses into strategies,
test them, and compound what survives.
Our endgame: human reasoning in
markets, a $1tn opportunity.

Turning compute into capital

Alpha Research — Orchestrator 8 AGENTS ACTIVE
EXPERIMENTS 38,471
UNIQUE STRATEGIES 12,041
BEST SORTINO 4.84
PORTFOLIO SORTINO 6.12
SESSIONS 142,910

Agents that research, test, and improve.

Our agents run inside a shared context. They build on prior work, test new hypotheses, and are learning continuously.

ORIENT
14,291 session logs · 12 lesson files · target: strategy_d4c71e8
Loaded an equity momentum strategy at Sortino 1.22. Prior sessions flagged horizon spacing as the bottleneck: lookbacks were too coarse in volatile regimes.
HYPOTHESIZE
4 candidate strategies · 2 new features designed
A: tighter geometric spacing. B: volatility-weighted spacing. C: ultra-short regime filter. D: PCA-residualized decomposition. Each adds a multi-horizon vol ratio and order-flow imbalance z-score.
EVALUATE
4 screened → 2 promoted to full backtest
C and D failed drawdown and IC gates on coarsened six-month data. A and B advanced to three-year backtests with regime splits.
candidate_A SR 3.18 · DD 2.7% · Calmar 31.4 PASS
candidate_B SR 0.9 · DD 2.6% · Calmar 8.1 FAIL
SYNTHESIZE
Sortino 1.223.18 · committed as strategy_d4c71e8_v2
Tighter spacing fixed the horizon issue. Lessons on horizon-ratio sensitivity and feature-pair constraints committed.
6 of 7 research tasks closed. Next: cross-asset correlation decay.

Incumbents scale headcount.
We scale agents.

A research lab where AI progress compounds directly into P&L.

Team from Citadel, Jump Trading, Stanford, Caltech, Berkeley

Founding research and infrastructure roles.