Senior Machine Learning Performance Engineer - Physics

NVIDIA • Santa Clara, CA

1d

About The Position

NVIDIA’s deep learning and HPC platforms have made a huge impact in various fields and are broadly used across leading academic institutions, start-ups, and industry, including the world’s largest Internet companies. We need passionate and creative people to help us building and optimizing an AI framework that will solve the toughest and most relevant problems of humanity and problems that are at the cutting edge of science & engineering: weather/climate challenges, product design, digital twins, molecular dynamics, novel materials, accelerated drug development, etc. What you'll be doing: Work with some of the brightest minds in a leading AI company to develop a leading machine learning framework, NVIDIA PhysicsNeMo, for our academic and industrial partners to construct digital twins and machine learning simulation surrogates for real world science and engineering problems Work with internal project teams to validate applications built using the framework on Nvidia’s products Characterize and optimize scientific AI workloads on the latest NVIDIA platforms. Stay up to date with the latest research and innovations in deep learning techniques, implement and experiment with new ideas to develop and enhance NVIDIA's deep learning technologies with focus on simulations

Requirements

• BS or MS degree (PhD preferred) in computer science, mathematics, computational science/engineering, or related technical field or equivalent experience
• 5+ yrs of relevant experience.
• Strong Python programming skills
• Familiarity with containers, numeric libraries, modular software design, distributed computing, and/or high-performance storage systems.
• Good knowledge of state-of-the-art DNN architectures and machine learning techniques and algorithms (graph networks, diffusion models, reinforcement learning etc.) with experience in developing or using major deep learning frameworks (PyTorch, Tensorflow, JAX etc.)
• Experience with solving and using machine learning for real world problems involving scientific/engineering simulations (domains/applications - industrial, life sciences, high energy physics, earth sciences – seismic, weather & climate modeling; physics types - CFD, structural, electromagnetics, optics, acoustics etc.)
• Experience with performance measurement tools, such as NVIDIA Nsight Systems
• Strong analytical skills with bias for action.
• Good time-management and organization skills to thrive in a fast paced, dynamic environment
• Solid written and oral communications skills
• Good teamwork and interpersonal skills

Nice To Haves

• Work with multi-node systems with data-parallel and model parallel programming experience.
• Experience with CUDA, or python kernel languages such as Triton or NVIDIA Warp
• Usage of nonlinear simulation tools and techniques, usage of major simulation codes (opensource and/or commercial).
• Experience with High Performance Computing for Scientific AI, or end-to-end training and inference optimization.
• Published papers in the field of AI in scientific computing
• Experience with scientific visualization is a big plus.

Responsibilities

• Work with some of the brightest minds in a leading AI company to develop a leading machine learning framework, NVIDIA PhysicsNeMo, for our academic and industrial partners to construct digital twins and machine learning simulation surrogates for real world science and engineering problems
• Work with internal project teams to validate applications built using the framework on Nvidia’s products
• Characterize and optimize scientific AI workloads on the latest NVIDIA platforms.
• Stay up to date with the latest research and innovations in deep learning techniques, implement and experiment with new ideas to develop and enhance NVIDIA's deep learning technologies with focus on simulations

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level