The Press
Specialty coffee aggregator pulling beans from 50+ independent roasters.
Engineering leader at Filevine. I publish essays occasionally — usually on management, teams, and how software actually gets built — and keep working notes more often. On the side I'm building The Press (a specialty coffee aggregator) and Pantheon (a CLI for coordinating LLMs), and tinkering with Eurorack synths when I'm not at a keyboard.
A personal newspaper, curated each morning by an LLM pipeline from sources I follow.
For past editions, see Archives →
A central system that computes, stores, and serves ML features so training and serving use identical values.
A proxy layer in front of model backends that handles routing, caching, rate limits, and fallback for LLM inference.
The platform layer above the GPU: the connective tissue that turns a model into a reliable production system, and the live shift from the feature-store MLOps stack to the vector-store LLM stack.
Specialty coffee aggregator pulling beans from 50+ independent roasters.
Terminal AI coding assistant supporting multiple LLM backends.
A curated list of open source Eurorack and synthesizer projects.