← all work

Autonomous AI trading agent

Nova

An autonomous AI trading agent, live in production. It scores setups, places trades, manages its own risk and logs every decision it makes.

Live
Overview. Live in production, trading a demo account.
Overview. Live in production, trading a demo account.
Insights. Nova reviews its own results and flags what to change.
Insights. Nova reviews its own results and flags what to change.
Brain. How the AI's grade compares with the rules checklist, and whether its confidence holds up.
Brain. How the AI's grade compares with the rules checklist, and whether its confidence holds up.
Analytics. Cumulative R and win rates by instrument, session, grade and direction.
Analytics. Cumulative R and win rates by instrument, session, grade and direction.
Risk gate. Four checks every trade has to pass before it's placed.
Risk gate. Four checks every trade has to pass before it's placed.

The problem

Good trading rules are easy to write down. Following them every single time, without emotion, is the hard part.

What I built

Nova is an agent that trades on its own, inside rules I set.

  • scores each setup with AI, using Smart Money Concepts analysis
  • executes trades and manages risk automatically
  • manages the portfolio as a whole, not trade by trade
  • live dashboards so I can see what it’s doing at any moment
  • a full audit trail of every decision and why it made it

Why the audit trail matters

An agent you can’t question is an agent you can’t trust. Every call Nova makes is recorded with its reasoning, so I can review it the same way I’d review a person.

How it’s built

  • Python and FastAPI
  • Claude API for setup scoring
  • Docker, running 24/7 on a VPS

My role

Designed, built and deployed solo. Execution logic, risk management, dashboards, infrastructure.

Where it is now

Live in production, placing trades on a demo account.