A collection of engineering deep-dives, architectural post-mortems, and performance benchmarks from building distributed web apps, canvas engines, and AI pipelines.
Decoupling stream extraction and transcode jobs using Redis background task queues and multi-worker audio piping to achieve sub-second playback startup.
Preventing React reconciliation bottlenecks during 60FPS physics updates with decoupled Zustand state stores and isolated requestAnimationFrame loops.
Architecting sub-second voice agent interfaces combining Deepgram streaming STT, Gemini 2.5 Flash function calling, and Google TTS.