Software engineer — distributed systems, data infrastructure, and production LLM products. Currently going deep on coding agents and local LLM inference on Apple silicon, in public.
Previously: co-founder of Aeqium (compensation platform, seed → ~$1.5M ARR · Braze, Whoop, Warby Parker); staff SWE at Day AI (Sequoia-backed AI-native CRM); computer-vision serving & training infra across 200+ stores at Walmart's Intelligent Retail Lab.
Coding agents & evidence tooling
- one-punch — spec-to-shipped pipeline for coding agents: intent capture, decision tickets, model-routed worker harness, artifact-verified completion
- evidence-kit — graded, source-mirrored research corpora for agents; evidence strong enough to rest decisions on
- outrigger — experiments in evidence-grounded machinery for long-horizon coding agents
- etl-compiler — spec-compiled ETL: auditable decisions, deterministic Python, quarantine over silent coercion
Local LLM inference on Apple silicon
- slipstream — 35B MoE on a MacBook in ~2 GB RAM: SSD-streamed experts, near-roofline Metal kernels, persistent KV cache
- gpu-kernel — measurement-first Apple GPU kernel work; decode attention at >90% of the memory-bandwidth ceiling; three optimizations merged upstream into Apple's MLX
- Writeups: dwijen.com
Elsewhere: drove the open-source release of walmartlabs/cookie-cutter, an event-driven TypeScript microservices framework.