Data Engineer with a strong interest in machine learning. I build scalable data systems by day and tinker with ML projects by night. Co-author on PACO (CVPR 2023). B.S. in Electrical Engineering from Sonoma State University. Based in the Bay Area.
I write about all of it on my blog.
- The Lottery Ticket Hypothesis — Reproducing and extending the lottery ticket hypothesis in JAX, with experiments on pruning, learning rate rewinding, and early-bird ticket detection
- Dispatch from the Frontier — Adventures in autonomous AI engineering with Claude Code
- Exploring Feature Stores with Apache Iceberg — Building a feature store with time-travel querying
- GPT-2 in Jax — Implementing and training GPT-2 from scratch
- Why Pre-Commit Hooks Are Essential — Setting up pre-commit for humans and AI agents
- Training a Neural Network with Jax — Building and training a neural network from scratch
- gitgym — Rustlings-inspired CLI tool for learning git through 32 hands-on exercises. Built almost entirely through autonomous AI coding with Claude Code.
- Feature Store with Iceberg — Feature store on Apache Iceberg with time-travel querying, snapshot-based model auditing, and anomaly detection debugging. Uses PyIceberg, DuckDB, and Docker.
- GPT-2 in Jax — Full GPT-2 implementation using Jax, Equinox, and Optax. Multi-headed attention, transformer blocks, training loops, and text generation with temperature/top-k decoding.
- Stock Bots — Three Bluesky bots posting real-time market data for U.S., international, and futures indices. Deployed with AWS Lambda and macOS launchd.
- Frugal AI Challenge — Audio classification for detecting illegal deforestation, built in Jax. Focused on efficient inference with mixed-precision training and high recall.
- UK AI Safety Institute - CORE-Bench — Extended the Inspect framework to test AI agents’ ability to reproduce 90 research papers across scientific domains.
- Minari RL Library — Fixed bugs, improved documentation, and developed tutorials for better library usability.
Python, Rust, SQL, Go | Jax, PyTorch, Equinox, Optax
AWS (Lambda, Glue, EMR, S3, EventBridge) | Spark, Iceberg, Databricks, Snowflake
Docker, CI/CD, Infrastructure as Code