Lead, AI Inference Optimization

DigitalOcean; Seattle (WA); Hybrid; USD 250,000 - 312,000; Full time

Job summary

DigitalOcean in Seattle is seeking a Principal Engineer to own model optimization across our accelerator fleet, including NVIDIA and AMD GPUs. You will define quantization strategies, kernel selection, and parallelism layouts to keep models fast on every GPU we buy.

You will lead end-to-end performance work, collaborate with NVIDIA/AMD teams, mentor engineers, and represent DigitalOcean in upstream communities. This hybrid Seattle role offers strong compensation and equity.

Qualifications

• 12+ years in performance-critical systems with experience optimizing LLM inference.
• Deep understanding of GPU architecture and the inference performance model.
• Hands-on kernel-level experience on CUDA/CUTLASS/Triton and ROCm/HIP/CK.
• Familiarity with serving engines such as vLLM, SGLang, TensorRT-LLM.
• Strong Python and C++/CUDA skills with profiling tools.

Responsibilities

• Set the technical strategy for model optimization across the fleet.
• Own quantization end-to-end including calibration and accuracy budgets.
• Drive performance across modern architectures and memory movement patterns.
• Lead speculative decoding work and acceptance-rate tuning.
• Write and tune kernels where upstream falls short and decide when to use them.
• Establish repeatable parallelism layouts per model and GPU family.
• Build benchmarking/regression infrastructure with meaningful data.
• Collaborate with NVIDIA/AMD teams and represent DigitalOcean publicly.

Skills

Performance optimization

GPU architecture

Python

C++/CUDA

Profiling & benchmarking

Cross-functional leadership

Tools

CUDA/CUTLASS/Triton

ROCm/HIP/CK

Nsight/rocprof

Job description

DigitalOcean in Seattle is seeking a Principal Engineer to own model optimization across our accelerator fleet, including NVIDIA and AMD GPUs. You will define quantization strategies, kernel selection, and parallelism layouts to keep models fast on every GPU we buy.

You will lead end-to-end performance work, collaborate with NVIDIA/AMD teams, mentor engineers, and represent DigitalOcean in upstream communities. This hybrid Seattle role offers strong compensation and equity.