Alex Wu

I creatively build ML systems to solve unorthodox problems. Recent NYU CS + Data Science grad.

Lorenz Attractor
Featured
1st Place — Datadog Self-Improving Agents Hackathon

CRUCIBLE

Two identical LLM agents play Split or Steal. No pre-programmed betrayal. Through private reflection, they discover deception autonomously. 86% mutual destruction rate. Peak entropy 1.00.

DevpostGitHub
Projects

Majikari

Japanese collectibles price comparison across marketplaces. YC S26 applicant. Live product serving real users.

View Project →
Startup

PsychohistoryML

Exploratory ML analysis of 372 historical polities from the Seshat Global History Databank. Random Forest (CV AUC 0.66) finds the complexity-duration relationship reverses by era.

View Project →
Research

Suzerain

Analyzes your Claude Code logs to reveal how you govern AI assistants. Bash acceptance rate is the key signal.

View Project →
CLI Tool

McCarthyGPT

Reverse-engineered Cormac McCarthy's prose style. 81% monosyllables create the "drumbeat" rhythm. 4.8M parameter character-level GPT trained on Blood Meridian.

View Project →
Deep Learning

Beyond Translation

Sentence embeddings reveal translator fingerprints across 5 English translations of Nietzsche. Hollingdale sits at the semantic centroid; Zimmern drifts furthest.

View Project →
NLP

Audio Style Classifier

89.5% accuracy on 954 Touhou tracks. Handcrafted audio features beat CLAP/MERT by 19-24%.

View Project →
Audio ML
Contact

Best way to reach me: amadeuswoo@proton.me