Operate production inference · 01
Reliability begins as product language
LatencyAvailabilityQualityData boundary
INFERENCE FOUNDER · FIELD COURSE
Operate production inference · 02
Release, capacity, request, incident
Safe changeEnough headroomPredictable executionFast recovery
INFERENCE FOUNDER · FIELD COURSE
Operate production inference · 03
Warm, burst, and batch are different products
Warm interactiveElastic burstInterruptible batch
03
INFERENCE FOUNDER · FIELD COURSE
Operate production inference · 04
A GPU disappears at peak traffic
Detect health lossStop new assignmentsProtect deadlinesRestore + reconcile
INFERENCE FOUNDER · FIELD COURSE
Operate production inference · 05
Practice the failure before the pilot
DetectProtectRecoverLearn
INFERENCE FOUNDER · FIELD COURSE