A quant desk, rebuilt as agents under your command.
You have a live demo in hand. This is the one-page version: who we are, why we built it, how it works, and what we'd like from you.
The infrastructure we needed didn't exist — so we built it.
We started as a DeFi-native quant fund, and the infrastructure we needed didn't exist. What was built for retail couldn't carry an institutional book: execution was inefficient, risk management an afterthought, and nothing was designed for AI agents. So we built our own.
That stack is now the product. It runs an AI-first book on a strict, transparent risk model, and it's produced double-digit yield over the last 18 months — all without a fifty-person desk.
Three forces are converging — and the on-chain infrastructure to ride them doesn't exist yet.
TradFi has arrived
Every major trading firm now runs a crypto arm — DRW's Cumberland, Jump Crypto, Susquehanna, Jane Street, Citadel Securities — and the frontier is moving from spot into DeFi.
The assets are following
Tokenized RWAs passed ~$30B on-chain in 2026, roughly 4× a year earlier. Forecasts run from ~$2T by 2028 (Standard Chartered) to ~$16T by 2030 (BCG).
The next traders are software
IDC projects 1.3B AI agents in workflows by 2028; agentic spend already hit ~$50B in 2025 (KPMG). Those agents need the same research, risk and execution foundations every trader relies on.
Anyone running, or wanting to run, strategies on-chain.
Curators
Build, risk-gate and run vault strategies with a defensible model and a full audit trail.
Hedge funds
The discipline of an on-chain quant desk — without having to build one.
Market makers
Systematic on-chain execution and risk gating.
The edge is yours. The AI does only what humans can't at on-chain speed.
No LLM picks a trade, sets risk appetite, or invents alpha — the thresholds are yours and the signing rules are deterministic.
Research
Collapses idea-to-backtested-strategy from weeks to minutes, so a researcher tests more hypotheses, not fewer.
Risk
Watches every position at block resolution. Humans sample a dashboard every few minutes; a liquidation can wipe a position in one block.
Execution
Fires a multi-leg atomic trade the second a window opens, and unwinds on breach — no waiting for two humans to sign.
Each agent maps to a tab in the live demo.
Spot the edge
Live spreads and signals across tokens, DEXs and lending venues, every chain — ranked the moment they appear.
→ Monitoring tab · LiveProve it
Backtests on real history with real frictions — slippage, gas, liquidations. The P&L you see is the P&L you'd have lived, with Sharpe, drawdown and regime breakdown.
→ Backtesting tab · LiveBuild it in plain English
Describe the trade; the agent writes the code, tests it and sharpens it with you. Work in our SDK or your own Jupyter. Your alpha never leaves your hands.
→ Strategy Build tab · LiveGate it — Odyssey Risk Model
Live technical and economic risk on every position and its underlying. Stress-test the book against any scenario, on a model whose weights your team tunes itself — built with a research chair at Institut Polytechnique.
Premium · On our own bookFire on your terms
One sentence becomes a fully-built, leveraged trade. Fire it yourself with a dry-run safety net, or hand it to a policy-bound co-signer that turns "two humans to sign" into "one human and an agent."
→ Execution tab · Paper / simTrack and report
Live P&L, equity and risk per fund on a private dashboard — then committee-ready performance, risk and attribution reports, with TCA on every fill to prove best execution to your IC and LPs.
→ Live P&L tab · LiveEvery module runs at the level you choose — and you move up the dial as trust builds.
You act
Odyssey as a classic app: it surfaces and computes, you act. TradFi-terminal parity.
Human-in-the-loop
Pre-built, automated actions — but every execution waits for your sign-off.
Runs in rules
Unattended inside fixed rules and limits. No per-action approval — and no judgment; it stops at the guardrails.
Judgment, gated
The agent reasons, adapts to new data and decides within policy. Still risk-gated, still under human command.
The agents come in three seniority tiers — seniority is how high an agent can climb the dial:
Assistant
Monitoring, reporting, alerts; watches, never trades.
Ceiling: Semi-autoTrader
Researches and builds strategies; you validate before anything goes live.
Ceiling: Full autoVP
Runs a full portfolio within its mandate: allocates, executes, rebalances, reports.
Ceiling: AgenticA week of DeFi engineering, from one line.
The fleet backtests it on the spot, then runs it around the clock — firing each opportunity as a single atomic transaction: flash-borrow WETH, swap into weETH, post Morpho collateral, draw the loan, and unwind the block if any leg breaches your limits.
It keeps backtesting as new data lands, and you fine-tune it in plain English against deep performance and risk metrics.
Three things.
- Test the demo.Build a strategy, backtest it, run it in paper.
- Try it with real capital.We'll set up your own Safe and take a strategy live (on request).
- Tell us what you think.Share your feedback with the team.
Like what you see? Once you're satisfied, we can onboard you directly — giving your desk first access to new arbitrage opportunities the moment they go live.
Where we're taking it.
- Expand the strategy library across new asset classes — stablecoins and tokenised RWAs, including tokenised equities and T-bills.
- Deepen the systematic playbook — basis and funding-rate arbitrage, cross-venue and cross-DEX arbitrage, delta-neutral liquidity provision, and statistical arbitrage.
- Complete the agent fleet — production co-signer, data, and reporting agents.
- Kick off our pre-seed round.