AI Compass
Compass
LEARNING PATHS

IN DEPTH

Hardware and operations

Why graphics cards, what memory bandwidth costs, and how a model reaches production.

about 4 hours

10 articles

  1. 01Why GPUs2 min
  2. 02GPU architecture2 min
  3. 03Memory and bandwidth2 min
  4. 04Training versus inference2 min
  5. 05Quantisation2 min
  6. 06Speeding up inference2 min
  7. 07Serving models2 min
  8. 08MLOps2 min
  9. 09Energy and sustainability2 min
  10. 10Edge AI · optional2 min
AI OPERATING PLATFORM

You don't have to know all of this yourself.

This wiki explains what needs doing. The LumeSec platform does it: it brings together what your AI systems are doing, holds them inside the agreed limits, and records the evidence as it goes.

See the platform
  • CorrelateOne picture of what your systems are actually doing.
  • ContainThe agreed limits hold at runtime, not just on paper.
  • AttestEvidence accrues in operation, not the week before an audit.
Hardware and operations