Senior MLOps engineer working on LLM training infrastructure, agent governance, and the question of whether LLM agents can sustain coherent research direction on their own.
- Synth — LLM-driven neuroevolution research. Six research streams run under university-style mentor guidance; 99.7% MNIST on ~17% of a dense baseline, plus a methodological finding that markdown lab notebooks suffice as cross-session memory for autonomous LLM research. Writeup.
- Periscope (internal) — Rust-based MCP server, 91 tools, ~170K embeddings over a curated philosophical-theological corpus.
- Carapace (internal) — RBAC sandbox for OpenClaw-class autonomous agents inside Gloo.
Rust, Python, PyTorch, Claude Code. Comfortable across the layers — from low-level Rust acceleration of PEFT init to research-direction supervision of agentic experiments.
Northwestern '18 (BS/MS CS, Murphy Scholar). Procedural content generation, Lisp/Racket worldbuilding projects, NLP research building on Google's Perspective API. Most of that lives in the older repos below.