Hi, I'm David.

I've spent the last twelve years building the unglamorous middle layer of modern AI: the pipelines, platforms, and serving infrastructure that turn promising models into systems a business can rely on. My work tends to live where large language models, vector search, and enterprise data meet — and where the failure modes are interesting.

Most recently I've been designing LLM-powered profiling and enrichment platforms — things like a Milvus-backed business intelligence system served via Triton, custom NER models trained on millions of documents, and a Spark rules engine that processes seven billion medical records on a schedule.

Before ML took over my calendar, I built data platforms (Snowflake, DBT, Airflow), shipped consumer-facing apps, and ran infrastructure on AWS, GCP, and Kubernetes. That history shows up in how I work: I sweat the boring parts — observability, cost, security, RBAC — because that's usually where ML projects die.

Outside of work I tinker with home automation, mess around with synth patches, and spend as much time as I can in the mountains around Salt Lake.