Senior AI Inference Optimization Engineer

DigitalOcean; Austin (TX); On-site; USD 191,200 - 239,000; Full time

Job summary

DigitalOcean is seeking a Senior Engineer 2 to drive architectural decisions that maximize throughput and minimize latency for large-model inference. You will lead benchmarking, performance optimization, and advanced kernel tuning across multi-node GPU clusters, working with CUDA, ROCm, and Triton toolchains.

You will mentor engineers, collaborate with product teams, and push for cutting-edge quantization and precision techniques to double throughput without sacrificing accuracy.

Qualifications

• 5+ years in high-performance computing or AI infrastructure.
• Deep familiarity with Gen AI landscape and major model families.
• Hands-on optimization of attention layers and parallelization across distributed GPUs.
• Comprehensive understanding of NVIDIA/AMD GPUs and software stacks.

Responsibilities

• Lead technical strategy for benchmarking and performance optimization at inference engine and GPU kernel layers.
• Engineer solutions for attention, memory, precision management, and multi-node GPU clusters.
• Implement cutting-edge optimization techniques to stay at the forefront of Gen AI.
• Mentor through code reviews and up-level engineering practices.
• Collaborate with Product Management to translate hardware limits into shippable features.
• Engage with GPU communities and contribute to open-source projects.

Skills

Technical depth

Gen AI literacy

Optimization expert

Hardware fluency

Open source mastery

Systems design

Leadership through influence

Low-level mastery

Triton/CUDA

Tools

CUDA

ROCm

TensorRT

OpenAI Triton

Job description

DigitalOcean is seeking a Senior Engineer 2 to drive architectural decisions that maximize throughput and minimize latency for large-model inference. You will lead benchmarking, performance optimization, and advanced kernel tuning across multi-node GPU clusters, working with CUDA, ROCm, and Triton toolchains.

You will mentor engineers, collaborate with product teams, and push for cutting-edge quantization and precision techniques to double throughput without sacrificing accuracy.