Services
I work best on backend problems that are expensive to get wrong: systems that move data, have to stay up, and need someone to own them end to end.
Event & Data Pipelines
Streaming ingest, transformation, fan-in/fan-out routing, backpressure, and bounded buffering. Kafka, Postgres, object storage, and the glue between them. I wrote LogBus because the off-the-shelf options were heavier than the problem.
Observability & Telemetry Cost
Audit what you collect, what you pay to keep, and what you actually look at. Then reshape the pipeline so the noise gets dropped early and the signal survives. See Telemetry Feeds, Kinetic Signals.
Search & Relevance
Elasticsearch/OpenSearch and pgvector, BM25 and dense retrieval, fusion and reranking, with offline evaluation so you know whether a change helped. See Search Lab.
Agentic Workflows
LLM agents that work through your existing tools (issues, pull requests, CI) with scoped permissions and full transcripts, so you can see what they did and why. See FJX.
Infrastructure for Small Teams
Provisioning and configuration for fleets that don't need Kubernetes. Plain, reviewable, idempotent. See Simple Configuration Management with Bash.
Engagement Shapes
| Shape | Good For |
|---|---|
| Assessment | "Something's wrong and we're not sure what." Fixed scope, written findings. |
| Build | A well-defined component or service, delivered with tests and docs. |
| Embedded | Working alongside your team for a period on a larger effort. |