nwyin.com

• billwatch — cached, token-efficient CLI for US state legislative data; agent-ergonomic design following AXI principles
• irradiate — mutation testing for Python, in Rust
• tau — minimal coding agent harness
• hive — multi-agent coding orchestrator
• ruse.sh - Rust port of the charm.sh TUI ecosystem
• pycg-rs — static call graph generation for Python
• pycfg-rs — control flow graph generation for Python
• dw2md — extract DeepWiki docs into a single LLM-ready file
• takeoff-protocol — multiplayer AI crisis simulation
• slop-guard — detect AI writing patterns in prose

• Nuuly as a Sr. Machine Learning Engineer.
  • During my first two years, I developed services, pipelines, and simulations to help with inventory management and logistics processing. These projects scaled from ~30k customers to ~350k customers, and saved (estimated) ~millions in shipping costs.
  • During my last couple of years at Nuuly I took ownership of our recommendation system. Working with two other data scientists, we built a transformer-based recommendation model that improved offline recall and precision by ~50% for active subscribers, and gave us dramatic improvement for "stale" customers (who haven't rented in >2 months).
• Sharing Excess to help reduce food waste from grocery stores and restaurants