Software engineer building real-time distributed systems, backend platforms, and AI infrastructure.
Backend & platform | Real-time systems | Agentic AI & MCP | Observability | Java | Python | Go
I've spent 10+ years building systems that run inline, where being slow or wrong is felt right away. Most of that has been on a real-time decisioning platform used by 1,000+ banks, working inside a 100-millisecond budget. I care about correctness under load, making failures visible early, and turning repeated work into tools other teams can pick up on their own.
- ⚡ Real-time backend services - low-latency APIs and rule services in Java and Spring Boot, with heavier work like analytics kept off the hot path.
- 🤖 AI infrastructure - MCP servers that give agents permission-scoped access to production data, and agentic workflows with validation gates.
- 🔭 Observability - metrics, tracing and alerting that show problems in hours instead of weeks.
- 🧩 Full-stack platforms - React and TypeScript front ends, including large migrations done without downtime.
- 🔌 Enterprise MCP server - rule, KPI and reporting data exposed to AI agents as permission-scoped tools; adopted by 6 teams. Container cut from over 5 GB to about 500 MB with zero CVEs.
- 🛡️ Agentic remediation framework - approved fix patterns plus an agentic workflow with a validation gate; closed 130 of 130 security findings, with its patterns now standard across 40+ repositories.
- ⚙️ Rule platform modernization - rule lookup went from 8 minutes to 12 seconds across 800+ active rules.
- 📈 Observability platform - coverage across 8.34M+ monthly requests; detection of missed SLAs went from 3 weeks to under a day.
- 🧱 Frontend migration - 85+ pages moved from Angular to React micro-frontends with zero production incidents.
This was internal work, so there's no code to link. I'm always happy to walk through the design and the trade-offs.
- 🧭 claude-skills - small, tested Claude Code skills. First one: session-finder, which finds a past session and resumes it in 1 click.
- 🛡️ AI Commit Guardrails - AI commit assistant that catches secrets in staged changes and redacts them before the LLM sees anything, then writes the commit message. Java, tested in CI.
- 🐞 SpotBugs #4354 - merged fix for an
OS_OPEN_STREAMfalse positive in the Java static analyzer: closing a wrapper stream now counts as closing the stream it wraps.
- Measure before building. The right fix often comes from checking what's actually happening first.
- Make it repeatable. Once a problem shows up a few times, the job is to build the tool, not fix it again.
- Keep things reviewable. Small, boring changes ship; clever ones sit in a queue.
- 🎓 Guest Lecturer, Stanford University (2×, 2025–2026)
- 📄 Co-author, Monitoring Fraudulent Transactions at Merchant Level in Real Time Payment, TD Commons 2021
I'm open to conversations about backend and platform engineering, real-time systems, and AI infrastructure.