What if one AI could run a trade from first idea to final fill?
Backtest. Forward test. Execute. Report.
One model stack, no hand-offs.
RFUND.AI
Frontier models research, backtest, forward-test, execute and report across 50+ asset classes. We're paid only from the profit they make you.
The lifecycle
Most funds split research, testing, execution and reporting across teams and tools. Ours run as one continuous loop, and every stage feeds the next one's training data.
Models propose strategies and test them against 20 years of tick, options and fundamentals data — with survivorship bias, slippage and borrow costs modelled, not assumed away.
Survivors trade paper capital against the live book for weeks. Anything whose live behaviour drifts from its backtest gets retired automatically.
Graduated strategies get real capital, sized by a risk model that watches correlation across the whole book. Orders route via smart order routing to venues with the best fill.
Every fill, every decision and the reasoning behind it is written up in plain English and delivered to you daily. Auditable, exportable, no jargon required.
The model stack
We don't ask one model to do everything. Research, risk and execution are different problems with different failure modes, so each gets a model tuned for it and a second model checking its work.
Reads filings, transcripts, macro prints and order-flow anomalies, then writes testable strategy specs — in code, with stated assumptions.
Classifies the current market regime and sizes exposure accordingly. Trained on volatility clusters, not just returns, so it cuts risk before the drawdown, not after.
Reinforcement-learned agents decide how to work an order — venue, size, timing — to minimise market impact against a live limit-order-book simulator.
Markets
Breadth is a risk tool, not a marketing number. When equities and rates move together, a strategy in agricultural futures or volatility products keeps the book uncorrelated.
Every class below is traded live today; the number is how many instruments we cover within it.
How we get paid
There's no management fee and no platform fee. If the book doesn't make money in a quarter, we don't invoice. It keeps our incentives exactly where yours are.
Your capital sits in your own brokerage account under your name. We have trading authority, never withdrawal authority.
Charged quarterly, only on gains beyond your account's previous peak. Losses are recovered before we earn again.
Infrastructure
Co-located inference, deterministic replay of every trading day, and a change-control process where models — like engineers — ship behind feature flags.
Access
Tell us about your capital and constraints. We'll come back within two business days with a paper-trading window so you can watch the book before committing anything.
Accredited investors and institutions. Minimum account $250,000.