← Projects
TRADING · PAUSED

MyTradeBot (Sentinel)

MyTradeBot (Sentinel)

MyTradeBot (internally, "Sentinel") was an ambitious algorithmic trading system: 261 S&P 500 stocks, an ensemble of machine-learning models (an MLP combined with XGBoost), and 93 engineered features feeding its decisions.

Technically, it was the most complex thing I had built. Methodologically, it was the least disciplined — and that is the whole point. I built it heads-down, optimizing as I went, without a written specification of what the system was actually supposed to do.

So it drifted. Not with a crash or a bug, but quietly: the system slowly became something other than what I had intended, and because there was no spec to measure it against, I could not even name the moment it went wrong.

I paused it. But the lesson became the foundation for everything that came after — the spec-first methodology I now apply to every project. MyTradeBot is the cautionary tale that proves the thesis by its absence.

PythonMLP + XGBoostData Analysis

Paused. The lesson here became the foundation for the spec-first methodology.