General Compute
Model Bring-up Engineer / ML Compiler Engineer
San Francisco +1 · Hybrid · Full-time · Posted Sep 1, 2026
- Location
- San Francisco
NYC - Experience
- 5+ years in systems or ML systems.
About the job
Bring new model architectures up on the company’s ASIC by compiling, verifying, and optimizing them for production inference.
What you'll do
- Own model bring-up from reference weights to correct tokens running on the ASIC.
- Build agentic tooling to compile models, compare outputs with references, identify failures, and iterate.
- Debug compiler issues, verify correctness, and optimize model performance on the hardware.
What they ask for
- 5+ years in systems or ML systems, with depth in ML compilers, model porting/bring-up, or high-performance kernels.
- Experience getting an architecture you did not design to run correctly on a different target, including numerics debugging.
- Strong knowledge of modern LLM inference, including transformers, attention, KV cache, MoE routing, quantization, and batching.
- Experience working inside a compiler stack at the IR, lowering, and operator-coverage level; familiarity with agentic tooling.
Nice to have
- Experience with a non-NVIDIA accelerator at the compiler or model bring-up layer.
- Kernel-level experience in CUDA, Triton, or a vendor kernel language.
Tools
MLIR/LLVM, XLA, CUDA, Triton, TPU, Trainium/Inferentia, Tenstorrent, Groq
AI focus
Build agentic tooling to automate model bring-up, including compiling, running, comparing outputs with references, localizing failures, and iterating; bring frontier LLMs and other model architectures to production inference on an ASIC.
Model-written summary · the original posting is authoritative