Skills to demonstrate in AI infrastructure work
This is a capability map, not a live vacancies feed or a hiring guarantee. Build artifacts that show how you reason about real serving constraints.
Inference performance engineering
Show a baseline, memory/roofline model, trace, controlled optimization, and quality-preserving result. Explain when the optimization loses.
ML platform and serving engineering
Show reproducible deployment, model/version management, routing, autoscaling, observability, and rollback.
AI infrastructure reliability
Show a bounded overload design, failure drill, tenant isolation, cost model, and recovery runbook.
Applied AI systems engineering
Show retrieval/agent task quality alongside latency, permissions, token usage, and cost.