Rust ML Systems Engineer: Real-Time GPU Inference

Archetype AI Inc.; San Mateo (CA); On-site; USD 180,000 - 280,000; Full time

Job summary

Archetype AI Inc. is building the world’s first physical AI platform to bring AI into the real world, with Newton as the core model.

You will own the serving path for Newton and related multimodal models, driving GPU utilization and low-latency inference from kernel to service. You will work on Rust-native inference nodes, CUDA kernels, and an end-to-end serving stack, collaborating with researchers to translate architecture changes into production serving work.

Qualifications

• 6+ years of software engineering with ML systems, inference, or high-performance GPU computing.
• Must have owned a production serving path end-to-end, not just benchmarking.
• Expert in PyTorch with production-shipped models; anticipate slowdowns before profiling.
• Strong CUDA or equivalent GPU depth: memory hierarchy, occupancy, Nsight profiling.
• Rust or C++ alongside Python, Linux performance, production ops; daily Rust work expected.
• Ability to translate architecture changes from researchers into serving work.

Responsibilities

• Build and own model nodes in a Rust inference runtime: loading, warmup, batching, streaming, GPU memory pools.
• Optimize kernels and the GPU path: custom CUDA kernels, mixed precision, quantization, parity against research.
• Own inference routing and serving: streaming path API to GPU node, request batching, SLO latency & cost.
• Productionize research checkpoints: export, compilation, quantization, parity evals, rollout.
• Build observability for inference: latency histograms, GPU metrics, OOM signals, replayable traces.

Skills

6+ yrs software eng

End-to-end serving

PyTorch in production

CUDA depth

Rust or C++

Researcher collaboration

Tools

Python

Linux performance

Job description

Archetype AI Inc. is building the world’s first physical AI platform to bring AI into the real world, with Newton as the core model.

You will own the serving path for Newton and related multimodal models, driving GPU utilization and low-latency inference from kernel to service. You will work on Rust-native inference nodes, CUDA kernels, and an end-to-end serving stack, collaborating with researchers to translate architecture changes into production serving work.