A selection of things I've built — solvers, tools, and games. Source is on GitHub.
My MIT MEng thesis (EECS), built with MIT Lincoln Laboratory's Global Influence Model. I applied statistical and ML methods — Singular Spectrum Analysis, an autoencoder, and an LSTM — to ~30,000 news articles to surface patterns and anomalies in influence campaigns.
My Georgia Tech MS Cybersecurity capstone. I benchmarked leading LLMs (ChatGPT, Claude, Gemini) against static-analysis tools at finding vulnerabilities in an IP camera's code, then generated proof-of-concept payloads. The LLMs found more distinct bugs at a comparable false-positive rate, but struggled with consistency and context.
My run at IMC Trading's global algorithmic trading competition — round-by-round market-making and trading strategies for a simulated exchange.
A Georgia Tech ML4T (CS 7646) final project comparing a hand-built indicator rule set (Bollinger %B, RSI, MACD) against a Q-learning agent on JPM. The RL agent dominates in-sample but fails to generalize — a clean bias–variance illustration. Full report shared on request (kept off the public site per GT's academic-integrity policy).
A Texas Hold'em bot that estimates hand equity with Monte Carlo simulation to drive its betting decisions.
Simulates card counting over many blackjack trials to quantify how the running count shifts expected value.
Play the dice game Perudo against a bot that runs Bayesian updates on the hidden dice each round. Can you beat it?
A streamer-pick optimizer for ESPN fantasy basketball — daily add/drops to maximize roster output under league constraints.
An MIT 6.863J Natural Language Processing final project: a model that predicts a missing word from the surrounding sentence context.
An MIT 6.4132 final project: integrated planning and decision-making for an autonomous agent.
Automated solvers for the viral 82-0.com roster game — one guarantees a perfect season by controlling the game's randomness, the other grinds it out through fast retries. Includes the reverse-engineering behind them.