Self-Hosted Quant Research System
At 15, I lost US$600 on a single trade — real money to me. Instead of trying to chase it back, I wanted to make smarter, tested decisions rather than emotional ones. That's what pulled me toward fintech: using technology and finance together to build solutions that help the real economy — starting with my own.
So I started building — teaching myself how trading systems are actually tested and run. This is a Python backtesting engine that tests systematic strategies on stock indices and crypto, and runs as a live service on a Raspberry Pi I set up as an always-on home server — queryable anytime from my phone through a Telegram bot.
Runs 24/7 on a Raspberry Pi home server. A Telegram bot lets me pull strategy results and current paper-trading positions from anywhere — no laptop needed.
Multiple research-grounded strategies (time-series & dual momentum, Donchian breakout, MA crossover, trend filters). Realistic costs on every test (0.1% fees, 0.05% slippage). In-sample vs out-of-sample split plus rolling walk-forward validation, with no look-ahead bias.
Paper-trading only, hard-guarded against real-money trades. API keys kept out of the code.
To trust a strategy only after it survives data it has never seen — treating overfitting as the main enemy and staying skeptical of good-looking backtests. And to use AI as a leverage tool to scaffold the Python modules, while keeping the design decisions and judgment my own.
