Senior Model Bringup & ML Compiler Engineer

General Compute Inc.; New York (NY); On-site; USD 180,000 - 300,000; Full time

Job summary

General Compute Inc. seeks a senior IC to own bringup and validation of frontier LLM architectures on our ASIC. You’ll build and own the end-to-end bringup loop from reference weights to first correct tokens, working with a hardware partner.

Focus is correctness first, then optimization, with responsibilities spanning compiler internals, graph lowering, and high-performance kernels. You’ll ship a production-ready path that makes models run fast and correctly.

Qualifications

• 5+ years in systems or ML systems.
• Deep depth in at least one: ML compilers, model porting/bringup, or high-performance kernels.
• Strong knowledge of modern LLM inference: transformers, attention, KV cache, MoE routing, quantization, batching.
• Comfortable inside a compiler stack (MLIR/LLVM, XLA, or vendor graph compiler).
• Fluent with agentic tooling and proactive bringup.
• Self-directed; not assigned tickets.

Responsibilities

• Own model bringup end-to-end, from reference weights to first correct tokens on an ASIC.
• Build the agentic bringup loop: harness of agents that compiles, runs, diffs, and localizes failures.
• Work in the compiler space: graph capture, IR lowering, op coverage, kernel selection; fix lowering or fused-kernel issues.
• Own correctness before speed: verification harness and end-to-end evals to match reference.
• Then optimize: operator fusion, quantization, memory layout, batching, KV-cache behavior.
• Collaborate with hardware partner's compiler and runtime team to productionize models.

Skills

ML systems

Model bringup

Compiler stacks

LLM inference internals

Agentic tooling

Job description

General Compute Inc. seeks a senior IC to own bringup and validation of frontier LLM architectures on our ASIC. You’ll build and own the end-to-end bringup loop from reference weights to first correct tokens, working with a hardware partner.

Focus is correctness first, then optimization, with responsibilities spanning compiler internals, graph lowering, and high-performance kernels. You’ll ship a production-ready path that makes models run fast and correctly.