- Location
- San Francisco, CA, United States · Remote
- Compensation
- $80k – $150k • 0.5% – 1.5%
- Employment
- Full-time
- Experience level
- MID (2–5 years exp)
- Specialization
- Wellfound
- Source
- Wellfound
Responsibilities
Design scheduling and graph-optimization logic for how AI workloads get mapped onto constrained hardware
Build quantization and precision-scaling tooling as a core lever for performance gains
Own early IR/graph-pass work - narrow scope by design (one model class, one hardware target, tight iteration loop)
Work closely with the systems engineer (Job 1) — this is explicitly a two-person collaboration for the first phase, not a solo build
Bias toward shippable, measurable results over general-purpose architecture
Requirements
MLIR/LLVM exposure or equivalent compiler / graph-IR experience (production ML compilers, quantization toolkits, or strong research-to-production background all count)
Comfortable owning deep-tech scope against a short, concrete proof point rather than a long spec