Senior RL Post-Training Systems Engineer

NVIDIA; Indiana (PA); On-site; USD 150,000 - 190,000; Full time

Job summary

NVIDIA is seeking a skilled RL frameworks engineer to architect and build scalable post-training infrastructure. You will optimize the RL training‑inference‑rollout loop across GPUs, CPUs, and LPUs, and contribute to open‑source RL frameworks and distributed runtimes.

You will collaborate with researchers, software and hardware teams to ensure the platforms meet the needs of cutting‑edge AI research, deliver reliable performance, and support production‑scale workloads across diverse environments.

Qualifications

• MS or PhD in Computer Science, Computer Engineering, or related field (or equivalent experience).
• 5+ years of professional experience in distributed systems, HPC, or ML infrastructure.
• Strong proficiency in Python and C/C++.

Responsibilities

• Architect and build RL post-training infrastructure that scales from experimentation on a single GPU to production across thousands of nodes.
• Tune RL training‑inference‑rollout loops on GPUs, CPUs, and LPUs for performance where it matters.
• Improve open-source RL frameworks and distributed runtimes (e.g., VeRL, Miles, TorchTitan, Ray, Monarch).
• Collaborate with researchers and hardware teams to prioritize capabilities and deliver robust production platforms.

Skills

Python

C/C++

Distributed systems

High-performance computing

Communication

Education

MS or PhD in Computer Science/Engineering

Equivalent experience

Tools

PyTorch

Kubernetes

Ray

Monarch

TensorRT

Job description

NVIDIA is seeking a skilled RL frameworks engineer to architect and build scalable post-training infrastructure. You will optimize the RL training‑inference‑rollout loop across GPUs, CPUs, and LPUs, and contribute to open‑source RL frameworks and distributed runtimes.

You will collaborate with researchers, software and hardware teams to ensure the platforms meet the needs of cutting‑edge AI research, deliver reliable performance, and support production‑scale workloads across diverse environments.