I build backend systems where correct state changes survive failure: database transactions, idempotent workers, retry and recovery paths, and explicit handoffs to downstream services.
My public work spans C#/.NET and PostgreSQL reliability, Go realtime-data pipelines, Rust systems engineering, human-reviewed AI workflows, and realtime protocol experiments.
Backend reliability · Data consistency · Go realtime systems · Rust systems · Human-governed AI workflows
Open to remote backend / platform engineering roles and focused engineering engagements.
C# / .NET · PostgreSQL · NATS JetStream · Testcontainers
A test-backed reference implementation for the database-to-message boundary.
It makes transactional outbox/inbox behavior, idempotent local effects,
FOR UPDATE SKIP LOCKED workers, retries, dead letters, and bounded delivery
guarantees concrete. Integration tests reproduce and prevent a stale
failure-write race between concurrent workers.
Go · bounded concurrency · realtime-data pipeline · deterministic tests
A clean-room, offline case study for typed market-data ingestion, bounded queues, semaphore-limited workers, store-scoped duplicate handling, and deterministic time-series summaries. It is an engineering demonstration with synthetic fixtures, not a trading product or a production reliability claim.
Rust · local-first transcript QA · human review
A bounded prototype for reviewing existing subtitle transcripts. It preserves the source and evidence for each candidate, and produces reviewed output only after explicit human decisions. Experimental retrieval and ranking are kept separate from the authoritative review path.
Rust · numerical integration · deterministic artifacts
An offline renderer for inspectable light paths in a Kerr spacetime. Its CPU
f64 geodesic path, typed ray outcomes, and validation artifacts keep physical
results distinct from display presentation. It is a research renderer, not an
Interstellar reproduction or a realtime GPU product.
Rust · TypeScript / React · WebSocket / WebRTC research
An experimental music-systems foundation with deterministic musical state, transport-independent protocol boundaries, a React/WASM lab UI, and browser-performance evidence. It is technical evidence, not a production DAW or an established product thesis.
Some projects use substantial AI assistance. I keep engineering ownership inspectable by defining constraints, reviewing diffs, and tying claims to code, tests, or reproducible artifacts; model output is not accepted automatically.
Projects here vary in maturity; each README states its scope, evidence, and limitations directly.



