Archetype
Machine Learning Systems Engineer
San Mateo, California, US · Full-time · Posted Sep 24, 2026
- Experience
- 6+ years in software engineering, several of them in ML systems, inference, or high-performance GPU computing.
About the job
Own and optimize the production inference serving path for Newton and related multimodal models, from Rust runtime and GPU kernels through routing, observability, and rollout.
What you'll do
- Build and maintain Rust inference runtime model nodes, including loading, warmup, batching, streaming, and GPU memory pools.
- Optimize GPU kernels and inference serving for utilization, precision, latency, and cost; productionize research checkpoints.
- Build serving observability, including latency histograms, GPU metrics, OOM signatures, and replayable traces.
What they ask for
- Expertise in PyTorch and experience shipping models to production.
- Strong CUDA or equivalent GPU expertise, including memory hierarchy, occupancy, and profiling.
- Experience owning a production serving path end to end.
- Rust or C++ alongside Python, Linux performance, and production operations; readiness to work in Rust daily.
Nice to have
- Experience with Rust ML stacks such as candle or Burn, or comparable Rust GPU compute.
- Experience with custom kernels or compiler stacks, production quantization and mixed precision, multimodal serving, or high-performance serving stacks.
Tools
Rust, candle, Burn, CUDA, PyTorch, Python, Linux, Nsight
AI focus
Serve and optimize Newton and related multimodal models, including GPU inference, custom kernels, quantization, and productionizing research checkpoints.
Model-written summary · the original posting is authoritative