Skip to main content
AMD

Performance Engineer (Inference, Training & GPU)

AMD San Francisco, CA, United States Full-time 1 day ago
Embedded & Systems Software

Originally posted November 20, 2026 by the employer.

About the role

As a Performance Engineer, you will optimize models to train and serve as fast as the hardware allows, identifying and eliminating bottlenecks in kernels, serving paths, and training loops. You will work up and down the stack, from low-level tensor and kernel optimization to fleet-wide serving efficiency.

What you'll do

  • Optimize inference and serving end to end across latency, throughput, batching, caching, and scheduling for production scale.
  • Write and tune GPU kernels using CUDA or Triton for hot paths, driving kernel fusion, memory- and bandwidth-bound optimization, and low-precision (FP8/INT8) execution.
  • Optimize training throughput and GPU utilization through parallelism strategies, communication/compute overlap, mixed precision, and eliminating pipeline stalls.
  • Build performance models, profiling workflows, and observability across throughput, latency, cost, and utilization.
  • Own numerical correctness across precision, kernel, and hardware changes.
  • Partner with researchers to productionize models for serving and improve experiment speed and reliability.

What you'll need

  • Performance-engineering experience across profiling, roofline analysis, latency/throughput optimization, and root-cause investigation.
  • Deep GPU programming and optimization experience with CUDA and/or Triton, including kernel-level tuning and memory hierarchy optimization.
  • Hands-on experience optimizing inference and serving for large models, covering batching, KV/prompt caching, quantization, and sampling.
  • Hands-on experience optimizing training performance with parallelism, distributed communication, mixed/low precision, and GPU utilization.
  • Working knowledge of ML framework internals such as PyTorch and/or JAX, including compiler paths like torch.compile or XLA.
  • Strong proficiency in Python, with the ability to work in C++/CUDA, Rust, or Go.

Nice to have

  • Experience at an AI lab or ML-native company optimizing systems used directly by researchers and productionizing research code.
  • Depth in low-precision numerics, including FP8/INT8 quantization, mixed-precision, and detecting numerical regressions across hardware platforms.
  • Experience with distributed systems for large-scale training and inference, including collective communication (NCCL), interconnects (NVLink), and model/tensor parallelism.
  • Experience serving generative, diffusion, video, or 3D/spatial models beyond text LLMs.
  • Multi-accelerator experience with GPUs, TPUs, or Trainium, including vendor collaboration on accelerator capabilities.
  • Experience building performance-modeling and observability frameworks for GPU utilization and cost.

Skills: CUDA, Triton, GPU optimization, kernel fusion, PyTorch, inference

Market context
  • Employer posting date: 20 Nov 2026
  • First seen in our index: 20 Nov 2026 (our observation date)
  • Last confirmed on the employer's career site: 20 Nov 2026
  • Specialty: Infrastructure/Platform Software (GPU Software Stack)
  • Comparable active jobs: 345 at 25 employers, as of 20 Nov 2026
  • Related specialties (Software, Firmware & Systems family): 6902 active jobs at 180 employers

View all AMD jobs · California jobs

How we counted: postings read from employer career sites; dates are our observations unless labelled as the employer's. How we count

Apply now
San Francisco, CA, United States
On-site
Full-time
1 day ago

Share this job