Pacific Design/ artificial intelligence

section 12 · status: live · 5 entries · updated 2026-08-29

AI Infrastructure & MLOps

Models get the headlines; infrastructure decides what ships. Why GPUs won and memory rules, how one network trains on ten thousand chips, the per-token economics of serving, and the operational discipline — classic MLOps and the new LLMOps layer — that keeps any of it trustworthy.

live · continuous batching — requests join mid-flight, the GPU never waits

  1. The hardware layermatrix multiplies, memory walls, and very expensive plumbing
  2. Training at scaleone loss curve, ten thousand chips, months of not crashing
  3. Serving & inferencethe meter runs per token — engineer accordingly
  4. The ML lifecyclethe model is 5% of the system; this is the other 95%
  5. LLMOpsprompts are config, evals are CI, tokens are COGS