Architecting resilient distributed systems by balancing throughput, reliability, and the trade-offs that matter at scale.
Senior engineer with 10+ years of experience designing, operating, and evolving production Kubernetes platforms and distributed systems. I make architectural decisions with a clear understanding of the operational tradeoffs and take ownership of the outcomes. My current work focuses on large-scale Kubernetes platforms and AI infrastructure, while I continue to explore agentic AI to accelerate thinking for more intelligent, reliable, and impactful platform operations. Based in Vancouver.
Markdown supported · ⌘↵ to post
Edge Ops
Silent 502s often mean the upstream closed after proxy_read_timeout while the client still waited. I log $upstream_status and $request_time, then raise the timeout only for the slow route instead of globally.