I build reliable backend systems and practical AI workflows with Go and Python—with an emphasis on observability, privacy, explainability, and safe failure modes.
- Backend platforms that stay predictable under failure
- Applied AI systems with evidence, confidence, and human review
- Developer tooling that makes complex workflows easier to operate
- Privacy-aware services that keep sensitive data inside trusted boundaries
| Project | Real-world problem | Engineering highlights |
|---|---|---|
| IncidentIQ | Alert floods hide the incident that actually matters. | Go ingestion API, deterministic deduplication, Python evidence-backed triage, graceful degradation, CI and containers |
| InvoiceGuard | Manual invoice review is repetitive, while blind automation is risky. | Explainable risk scoring, duplicate detection, human-review thresholds, Go orchestration and Python extraction |
| RedactSafe | Sensitive data often leaks into logs, support tools, and AI prompts. | Self-hosted redaction API, exact finding locations, Luhn validation, no persistence, bounded requests |
| JudgeMyCode | Interview practice needs realistic execution, feedback, and progress tracking. | Go API, PostgreSQL, React/TypeScript, asynchronous execution, revision-safe autosave and GitHub Actions |
I contribute fixes upstream when a problem is best solved at the source. Recent work includes adding bounded incoming MQTT packet handling to Eclipse Paho's Go client, addressing resource-safety concerns for production consumers.
- Make failure visible and recoverable.
- Keep AI recommendations explainable and reviewable.
- Treat privacy, limits, timeouts, and auditability as product features.
- Ship documentation, tests, and a reproducible local setup with the code.
I'm interested in backend, platform, and applied-AI work where reliability matters. Explore the projects above or start with IncidentIQ.