Senior Deep Learning Inference Engineer - Distributed Systems

NVIDIA AI; Santa Clara (CA); On-site; USD 152,000 - 288,000; Full time

Benefits offered by this job

Equity

Benefits

Job summary

NVIDIA seeks a Senior Deep Learning Algorithms Engineer to advance Dynamo, our open-source distributed inference platform for large-scale, low-latency AI services. You’ll lead architecture and performance work across Dynamo and open source frameworks, collaborating with research, software, systems, and hardware teams to make AI inference faster, more efficient, and easier to deploy.

Responsibilities include designing and maintaining Dynamo integrations with vLLM, SGLang, and TRTLLM; reducing

Qualifications

• BS, MS, or PhD in Computer Science, Electrical Engineering, Computer Engineering, or related field (or equivalent).
• 3+ years building, profiling, and debugging performance-critical distributed or ML systems.
• Strong programming skills in Python and/or Rust, C++.

Responsibilities

• Design, build, and maintain Dynamo integrations for open source frameworks vLLM, SGLang, TRTLLM.
• Partner with open source communities to land measurable gains in latency, throughput, reliability, and efficiency.
• Showcase NVIDIA token/watt leadership by pushing the pareto frontier on public/private benchmarks
• Find and remove bottlenecks across runtimes, kernels, networking, routing, and orchestration.
• Develop inference optimizations for scheduling, disaggregation, KV caching, and autoscaling.

Skills

Python

Rust

C++

Education

BS/MS/PhD in CS/EE/related

Job description

NVIDIA seeks a Senior Deep Learning Algorithms Engineer to advance Dynamo, our open-source distributed inference platform for large-scale, low-latency AI services. You’ll lead architecture and performance work across Dynamo and open source frameworks, collaborating with research, software, systems, and hardware teams to make AI inference faster, more efficient, and easier to deploy.

Responsibilities include designing and maintaining Dynamo integrations with vLLM, SGLang, and TRTLLM; reducing