How it works

Not a prediction, an auditable process

Every FinAI signal goes through the same four-stage pipeline. No signal reaches you without passing adversarial validation and the rule filter.

We get to know you first

Risk profile & suitability

You complete a short questionnaire covering your time horizon, loss tolerance, financial capacity and goals.
Your answers become a risk score from 0 to 100 and a band (Conservative / Balanced / Aggressive).
That band sets the maximum equity and crypto weight your portfolio can tolerate, and filters every signal.

We test the strategy on the past

Pick a ready-made strategy (Buy and hold, SMA crossover, RSI, Momentum) or set your own parameters.
The engine runs a realistic simulation that includes commissions, slippage and Turkish withholding tax (stopaj).
Compare CAGR, Sharpe ratio, maximum drawdown and win rate against a buy-and-hold benchmark.
A walk-forward approach is the default safeguard against overfitting.

We try to knock the signal down

The generator model proposes an action (BUY / REDUCE / HOLD / SELL) with a confidence level and a rationale.
Three independent sceptical lenses then step in: risk, liquidity and macro.
If at least two lenses refute the signal, it is never shown to you.
Even rejected signals are logged for transparency — this improves accuracy and is evidence of our duty of care.

We keep everything in the open

Every signal that passes validation is written to the immutable log, chained to the previous entry with SHA-256.
If any entry is altered afterwards, chain verification breaks immediately.
The signal's realised outcome — gain or loss — appears on the public transparency scoreboard.
The portfolio guardian watches your positions and warns you about concentration and risk drift.

Signal pipeline

Generator model

Adversarial panel

Rule filter

Immutable log

The pipeline is built behind an interface so a real AI engine (e.g. Claude) can be swapped in; the validation and logging logic stays unchanged.

Try it now