I build things that break gracefully and scale boringly
fn my_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",
}
}π― Turn uncertainty into architecture β When "we think it scales" needs to become "we know it breaks at X"
π« Say no to complexity β Every dependency is a future incident
Languages
Stack
β Stopped a 6-month rewrite β Data access was the bottleneck, not architecture
β Constrained AI hallucinations β Confidence thresholds over blind answers
β Eliminated race conditions β Idempotency over horizontal scaling
β Optimized RAG pipeline β 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."
π§ Email β’ πΌ LinkedIn β’ π¦ Twitter
Most work in private repos. References on request.