Model Bring-up Engineer / ML Compiler Engineer
General Compute · San Francisco · NYC
Infrastructureremotesenior
At a glance
Highlights
- Own new frontier models from reference weights to first correct tokens in days.
- Build agentic harness that compiles, diffs, and localizes failures automatically.
- Make decode 5–7× faster by running workloads on purpose-built silicon.
Ideal candidate
A self-directed systems builder who loves landing new models on unfamiliar hardware and automating the grind away. You obsess over correctness first, then cost-per-token, and thrive owning ambiguous bring-up end-to-end.
Requirements
- 5+ years in systems or ML systems with depth in ML compilers, model porting/bringup, or high-performance kernels
- Strong understanding of modern LLM inference: transformers, attention, KV cache, MoE routing, quantization, batching
- Comfortable inside a compiler stack at the level of IR, lowering, and op coverage
- Fluent with agentic tooling and building automated bring-up loops
- Self-directed owner who drives bring-up end-to-end without assigned tickets
Lab profile
About General Compute
Neocloud for alternative chips, running fragmented inference on purpose-built silicon.
What they build
- Alternative-chip inference cloud
- Split prefill-decode inference serving
Known for
- Productionizing SambaNova, Cerebras, Positron, d-Matrix hardware
- Rack procurement and data center operations
- ML compiler and model bring-up
Posted · Last seen