What you'll do
- Design and own retrieval + router strategy across environments
- Wire eval harnesses as release gates; track win‑rate & regressions
- Profile/optimize inference: caching, quantization, prompt trims
- Partner with product to translate wins into measurable deltas
What we're looking for
- 5–8+ years in ML platform/infra; Staff‑level scope
- Deep Python; experience with PyTorch/Triton and serving stacks
- Hands‑on with vector DBs, observability, and CI/CD for models
- Clear written communication; comfort with PR reviews
Where this role fits
This role sits in the LLM Infra & MLOps TalentZone and commonly supports packs like Cost & Quality Sprint and Guardrail & Reg Audit. Work is routed under QA; this is a managed bench, not a marketplace.