Sina Nezhadian

Research

Fleet-scale quality: AI hardware reliability and customer quality

AI hardware now ships the way rail fleets do: thousands of identical units, in the field, with a customer who reports every failure. On BART’s 775‑car fleet I led 8D investigations and drove customer non-conformances to verified closure. This section applies those methods — 8D, CAPA, first-article inspection, test readiness and lessons learned — to the reliability problems AI hardware teams face now. I haven’t worked in a data center; each piece says where the rail lesson transfers and where it doesn’t.

Flagship studyHow the best tech companies find the real cause — and what it means for customer quality

Related: Productive Compute, independent research on turning purchased AI hardware into usable capacity.

Coming next

  1. 01What a 775‑car fleet teaches about AI hardware reliability · in progress
  2. 028D for accelerator and server field failures: a working template · in progress
  3. 03Customer non-conformances at fleet scale: from first report to verified closure · in progress
  4. 04First-article inspection for AI hardware: what to check before volume · in progress
  5. 05Test-readiness reviews: lessons from commissioning a driverless system · in progress
  6. 06Lessons learned that travel: applying REX across sites and suppliers · in progress

Published