NVIDIA
4 months ago
Deep Learning Engineer - Distributed Task-Based Backends 2 Locations
$148k - $276kNVIDIA
We are looking for Senior to Principal level experienced software professionals to help build the next generation of distributed backends for premier Deep Learning frameworks like PyTorch, JAX and TensorFlow. You will build on top of validated task-based runtime systems like Legate, Legion & Realm to develop a platform that can scale a wide range of model architectures to thousands of GPUs!
What You Will Be Doing:
Develop extensions to popular Deep Learning frameworks, that enable easy experimentation with various parallelization strategies!
Develop compiler optimizations and parallelization heuristics to improve the performance of AI models at extreme scales
Develop tools that enable performance debugging of AI models at large scales
Study and tune Deep Learning training workloads at large scale, including important enterprise and academic models
Support enterprise customers and partners to scale novel models using our platform
Collaborate with Deep Learning software and hardware teams across NVIDIA, to drive development of future Deep Learning libraries
Contribute to the development of runtime systems that underlay the foundation of all distributed GPU computing at NVIDIA
What We Need To See:
BS, MS or PhD degree in Computer Science, Electrical Engineering or related field (or equivalent experience)
5+ years of relevant industry experience or equivalent academic experience after BS
Proficient with Python and C++ programming
Strong background with parallel and distributed programming, preferably on GPUs
Hands-on development skills using Machine Learning frameworks (e.g. PyTorch, TensorFlow, Jax, MXNet, scikit-learn etc.)
Understanding of Deep Learning training in distributed contexts (multi-GPU, multi-node)
Ways To Stand Out From The Crowd:
Experience with deep-learning compiler stacks such as XLA, MLIR, Torch Dynamo
Background in performance analysis, profiling and tuning of HPC/AI workloads
Experience with CUDA programming and GPU performance optimization
Background with tasking or asynchronous runtimes, especially data-centric initiatives such as Legion
Experience building, debugging, profiling and optimizing multi-node applications, on supercomputers or the cloud
You will also be eligible for equity and benefits. NVIDIA accepts applications on an ongoing basis.