Orbis — Algorithmic Trading Platform
The idea behind Orbis is a self-correcting strategy loop: a trading idea almost never works on the first try, so it has to be backtested to find where it fails, corrected, and tested again — repeatedly, until it becomes consistently profitable. Orbis is built toward eventually automating that whole loop with AI agents that write a strategy, backtest it, read the results, and revise it on their own, then keep monitoring and adjusting it over time as conditions change. The codebase is already shaped for that future: strategies plug into the backtester and the live bot through a small registry instead of being hardcoded, so an agent that writes a new strategy just has to register it — nothing else in the system needs to change.
Orbis is a full-stack trading platform: a FastAPI backend connects to Binance and Capital.com, runs an event-driven backtesting engine against historical data, and can execute strategies manually or through running bots in real time. The backtester is tested against the mistakes that make backtests lie: no look-ahead (a signal only fills at the next bar's open, never the current bar's own close), and the Sharpe ratio annualized for the data's real bar spacing instead of a hardcoded trading-day count. The Next.js frontend shows live candlestick charts for Bitcoin, NASDAQ 100, and Gold, lets you backtest a strategy before risking anything, place manual trades, and monitor open positions and account balance.
Today the two included strategies (SMA crossover, RSI) are still written and tuned by hand — the self-correcting agent loop described above is the direction the project is being built toward.
A further idea for later: AI models answer based on probability learned from past examples, and the same structure could apply to price data — training a model on an asset's historical movements to estimate the probability that a pattern repeats, and trading on that probabilistic prediction.
Built solo: the backend trading engine, broker integrations, and the dashboard/backtesting frontend.