Most of my work is for clients, so the case studies describe each system by industry and leave out client names. The open-source projects link to their code.
Ingestion pipelines that chunk, embed and load documents into vector databases, with agent workflows built on LangChain, LangGraph and Google ADK for retrieval and automated insights.
Retrieval and reporting that analysts used to do by hand.
Split a large IoT backend into independently deployable services, with a Consul service mesh for discovery, health checks and traffic management.
Services ship on their own schedule instead of one release train.
Re-architected legacy Databricks jobs as versioned, tested dbt models with clear lineage between sources and marts.
Transformations became reviewable SQL with tests in CI.
An AI workflow that turns a reporting request into a GCP Looker dashboard, replacing a manual build for every new report.
Python emulated speakers that join rooms and stream audio at scale, used to find and fix Janus server failures under high concurrency.
Led the move from Redux-Saga to React Context and TypeScript, rewrote legacy screens, and cut the landing page bundle by 10x.
10x smaller landing page bundle.
Connected third-party clinical systems to an enterprise EHR over HL7 FHIR, and added SAML single sign-on for enterprise customers.
Integration delivery went from days to hours.
Integration tests that run at each stage of an AWS data pipeline, verify its artifacts and send a report when the suite finishes.
A researcher agent holds every tool and a deterministic calculator; a writer agent holds none and can only use figures the tools returned. Covers all 50 states and DC.
Plans three truck routes, checks the forecast at each checkpoint for the hour the truck arrives, scores wind, rain, snow and load risk, and recommends the safest route.
A command-line tool that exports every product, or selected collections, from a Shopify store to CSV.
Tell me what you’re building and what a good outcome looks like. Full-time roles are welcome too.