Senior High-Performance LLM Training Engineer
Core
Optimizing NVIDIA's high-performance LLM software stack (PyTorch, JAX) for training on thousands of GPUs and shaping next-gen GPU hardware roadmaps.
Role type
Senior High-Performance LLM Training Engineer (IC)
Builds
Production-quality software for NVIDIA's deep learning platform stack (drivers to frameworks), MLPerf Training benchmark submissions, and proprietary processor/system simulators.
Domain
AI/ML Infrastructure, GPU Computing, Computer Architecture
Deliverable
production ML models
Required skills
Deep learning & neural network training, GPU architecture fundamentals, C++, Python, CUDA, application performance analysis, system-level performance modeling, software implementation across stack layers.
Preferred skills
PhD in CS/EE/CE, experience with state-of-the-art neural networks, familiarity with DL frameworks.
Technologies
PyTorch, JAX, CUDA, C++, Python, MLPerf, NVIDIA proprietary simulators.
Responsibilities
Profile and optimize AI training workloads on innovative hardware/software platforms; solve performance problems across state-of-the-art neural networks; implement key DL training workloads in simulators; build tools to automate workload analysis and optimization.
Seniority
Senior, hands-on IC
