Short technical notes, links, lessons, and observations from production work.
Video AI Evaluation
Manual review of MSRVTT, MSVD, VATEX, DiDeMo, and QVHighlights shows cases where the benchmark ground truth is too narrow, shifted, or ambiguous, so valid retrieved clips are scored as misses.
VLM Evaluation
A small chessboard-to-FEN experiment showed Claude Opus 4.7 struggling with exact square-level localization even when it understood the board broadly.
Operations
A short field note on replacing repeated bare requests calls with a shared session so Python services can reuse HTTP connections.
Frontend Performance
A practical note on preflight caching, custom auth headers, and why OPTIONS requests quietly dominate app latency on slower connections.
Operations
A field note from a simple offline Postgres migration: why pg_dump was enough, where read-only mode fits, and why the restore drill matters more than the dump command.
Media Infra
Why API Gateway compression can make large responses look safe in testing, then still fail when Lambda enforces the raw payload limit first.
Build Systems
Slim bases, same-layer cleanup, CPU-only PyTorch, pinned wheels, and the unglamorous work of keeping deploys fast.
