FableBot

FableBot is an algorithmic modeling platform I built solo to push my systems and data-engineering skills on a genuinely hard problem. It ingests years of market data, searches for trading strategies, and validates them under strict statistical guards. It’s paper-trading only, with no real money and no LLM anywhere in the stack, which is the whole point.

What I built

• A streaming ETL pipeline that processes a 56 GB CRSP market-data set row-by-row in constant memory, applying split and dividend adjustments and emitting clean per-symbol data for backtesting.
• A composable strategy layer searched by a genetic algorithm and Bayesian optimization, plus gradient-boosted-tree and Q-learning models I implemented from scratch in NumPy.
• A walk-forward validation gate (out-of-sample testing, purged cross-validation, transaction-cost modeling) that keeps strategy selection and validation strictly separated, so a strategy can’t be tuned on the data it’s later judged against.
• The whole platform designed, built, tested, and operated solo, backed by 260 automated tests run as a pre-merge gate, with the safety guarantees encoded as executable invariants rather than comments.

Why it matters

Most of my other projects are full-stack apps with an AI service inside. FableBot is deliberately the opposite. There’s no web front end and no model behind an API here, only data pipelines, search algorithms, and the engineering that keeps their output trustworthy. The real work was in that last part: keeping strategy selection and validation strictly separate, and proving the risk rules hold on every bar. It’s the project I’d point to for systems and data-engineering depth.