Senior AI Inference Optimization Engineer

DigitalOcean; Seattle (WA); Hybrid; USD 191,000 - 239,000; Full time

Job summary

DigitalOcean is seeking a Senior Engineer 2 to lead architectural decisions for inference performance and to optimize GPU kernel and inference engine performance. You will solve memory bandwidth bottlenecks and guide the technical roadmap for a high-performance inference fleet.

You will collaborate with product teams and TPMs to turn theoretical hardware limits into shippable features while mentoring engineers and driving open-source collaboration in the GPU AI space.

Qualifications

• 5+ years of experience in high-performance computing or AI infrastructure.
• Strong familiarity with Gen AI landscapes and model families.
• Hands-on optimization experience for attention layers and parallelization.
• Deep understanding of NVIDIA/AMD GPU architectures and software stacks.
• Open-source software integration and contribution experience.
• Systems design with low-level GPU programming and memory access patterns.
• Leadership by influence with cross-functional collaboration.
• Proficiency with Triton or CUDA toolchains.

Responsibilities

• Lead benchmarking and performance optimizations for inference engines and GPU kernels.
• Engineer solutions for memory bandwidth and compute utilization bottlenecks.
• Develop and deploy FP8, BF16, FP4 quantization techniques.
• Identify kernel fusion opportunities and multi-node GPU parallelization.
• Collaborate with product teams to translate hardware limits into features.
• Mentor through code reviews and elevate technical standards.

Skills

High-performance computing

Gen AI literacy

Attention optimization

Distributed GPU

CUDA proficiency

ROCm proficiency

Triton

Open source

Tools

CUDA

ROCm

TensorRT

Triton compiler

Job description

DigitalOcean is seeking a Senior Engineer 2 to lead architectural decisions for inference performance and to optimize GPU kernel and inference engine performance. You will solve memory bandwidth bottlenecks and guide the technical roadmap for a high-performance inference fleet.

You will collaborate with product teams and TPMs to turn theoretical hardware limits into shippable features while mentoring engineers and driving open-source collaboration in the GPU AI space.