I build things that break gracefully and scale boringly
fnmy_approach() -> Philosophy{match project_phase {Early => "Make it work, make it right, then make it fast",Growth => "Measure twice, optimize once",Crisis => "Fix guarantees before adding servers",}}🎯 Turn uncertainty into architecture → When "we think it scales" needs to become "we know it breaks at X"
🚫 Say no to complexity → Every dependency is a future incident
Languages
Stack
✓ Stopped a 6-month rewrite → Data access was the bottleneck, not architecture
✓ Constrained AI hallucinations → Confidence thresholds over blind answers ✓ Eliminated race conditions → Idempotency over horizontal scaling
✓ Optimized RAG pipeline → Capped recall, added reranking, 40% cost reduction
- Distributed systems are guilty until proven innocent
- Types at compile time are free, at runtime they're expensive
- The best abstraction is the one you can debug at 3 AM
- "Just scale it" is rarely the right answer
- If it doesn't have a postmortem DB, it's not production-ready
"Most failures start as reasonable assumptions."
📧 Email • 💼 LinkedIn • 🐦 Twitter
Most work in private repos. References on request.