Member of Technical Staff — Inference-Multi-Hardware

What you'll do

• Bring up systems on new accelerators and port and optimize kernels across hardware and programming models.
• Design cross-vendor hardware abstractions and build benchmarking, profiling, and regression detection.
• Debug correctness and performance issues, collaborate with silicon vendors and internal teams, and contribute optimizations to SGLang and Miles.

What they ask for

• 4+ years of experience in systems, performance, or ML infrastructure engineering.
• Deep expertise in at least one accelerator programming model and ability to learn new ones quickly.
• Experience writing or optimizing high-performance ML kernels and with distributed execution and communication libraries.
• Proficiency in C++ and Python; strong system-level debugging and profiling skills; production-shipped performance work.

Nice to have

• Experience bringing up ML workloads on new silicon or working across multiple vendor ecosystems.
• Experience with compiler stacks, hardware abstraction layers, quantization, distributed inference or training systems, CPU inference, large-scale communication, open source, vendor collaboration, or HPC.

Pay

$200,000 - $400,000 USD (base salary)

Tools

SGLang, Miles, NVIDIA, AMD, Google TPUs, CUDA, ROCm/HIP, C++

AI focus

Performance engineering and portability for AI inference and training systems across accelerators, including kernel optimization and debugging numerical correctness.