Senior Inference Optimization ML Engineer
Core
Optimize and deploy large multimodal foundation models (vision, video, language, proprioception, action) for cost-efficient inference across diverse hardware, including AMD GPUs, Google TPUs, wafer-scale accelerators, and on-robot embedded systems.
Role type
Senior Inference Optimization ML Engineer
Builds
Cost-efficient inference pipelines for Rhoda's generalist intelligent robots
Domain
Robotics + ML Systems Optimization
Deliverable
production ML models
Required skills
PyTorch, model quantization (INT8/FP8/FP4, AWQ/GPTQ), model pruning and distillation, kernel authoring (Triton, Pallas, MLIR), hardware performance profiling, inference serving frameworks (vLLM, SGLang, Triton Inference Server), debugging and measurement
Preferred skills
Diffusion model optimization, experience with non-NVIDIA accelerators (AMD ROCm, TPU, Groq, Tenstorrent), Jetson/Thor-class embedded deployment, open-source ML compiler contributions, streaming/low-latency systems
Technologies
PyTorch, ROCm, HIP, MIOpen, XLA, JAX, Pallas, Triton, MLIR, TensorRT, vLLM, SGLang, JetStream, IREE, TVM, CUDA, FP8/FP4
Responsibilities
Port and optimize models for non-NVIDIA and lower-cost hardware; deploy and optimize models on Jetson/Thor-class embedded compute; build cost/performance models for hardware selection; write and tune hardware-portable kernels; apply model compression techniques; build benchmarking and regression infrastructure; work with research engineers to design efficient models
Seniority
Senior, hands-on IC
