We are seeking highly motivated and skilled systems engineers to join our team to help in developing an AI Platform that offers an efficient infrastructure for inference and training large scale models. As a systems engineer, you will play a crucial role in building a unified solution that brings our innovative NVIDIA technologies such as high-performance, inference/training frameworks, ML compilers, performance predictor, and cluster scheduler into a single, cohesive platform.
What you will be doing:
- Taking part in the development of the NVIDIA's AI platform for training, fine-tuning and serving latest and greatest AI models with the best performance and efficiency.
- Designing and building solutions for scheduling large scale AI training and inference workloads on GPU clusters over many cloud infrastructure.
- Exploring and finding solution for open problems like industry-scale resource management, GPU scheduling, performance prediction, and live workload migration.
- Work with and contribute to adjacent teams like TensorRT/Dynamo inference engine, ML compiler, KAI/Grove scheduler, Lepton cloud etc.
What we need to see:
- Bachelor's degree or equivalent experience in Computer Science, Computer Engineering, relevant technical field.
- Experience building large scale systems from scratch. Prior experience in container-based deployment systems like Kubernetes is beneficial.
- Strong coding skills in programming languages like Python, Go, Rust and/or C/C++.
- Solid foundation in other computer science and computer engineering topics: algorithms and data structures, operating systems, computer architecture, etc. Strong understanding of AI and related technologies is a huge plus.
- Most importantly, ability to quickly grasp new concepts and thrive in evolving situations.
Ways to stand out from the crowd:
- Graduate-level education or relevant practical background, particularly in research, is beneficial.
- Practical experience in building and optimizing AI applications is highly desired.
- Proficiency in container software such as containerd, CRI-O, Linux namespace, CRIU, and NVIDIA GPU technology such as CUDA graphs, Driver/runtime is greatly advantageous.
