Solutions Architect - Deep Learning

Position Overview

NVIDIA's Worldwide Field Operations (WWFO) team is seeking a Solution Architect with a strong focus on Deep Learning and a deep understanding of neural network training. The introduction of NVIDIA GB200 NVL72 systems, bringing Chip-to-Chip NVLINK and the significant expansion of the NVLINK domain, have enabled a wide range of new neural network architectures and approaches to training. The ideal candidate will be proficient using tools such as NeMo, Megatron-LM, DeepSpeed, PyTorch FSDP, or similar, and have strong systems knowledge, enabling customers to fully utilize the capabilities of the new Grace Blackwell training systems. This could include helping customers take advantage of much wider neural networks, use of asynchronous checkpointing, or activation offloading. Hands-on experience in LLM post-training, particularly RL, would help the candidate stand out.

Solutions Architects work with the most exciting computing hardware and software, driving the latest breakthroughs in artificial intelligence! We need individuals who can enable customer productivity and develop lasting relationships with our technology partners, making NVIDIA an integral part of end-user solutions. We are looking for someone passionate about artificial intelligence, who can maintain understanding of a fast-paced field, and is able to coordinate efforts between corporate marketing, industry business development, and engineering. Solutions Architects are the first line of technical expertise between NVIDIA and our customers. Your duties will vary from working on proof-of-concept demonstrations to driving relationships with key executives and managers in order to promote adoption of NVIDIA-based AI technology. Engaging with developers, scientific researchers, data scientists, IT managers, and senior leaders is a significant part of the Solutions Architect role.

What You Will Be Doing

• Work directly with key customers to understand their technology and provide the best AI solutions/guidance on training processes in terms of tools and methodology.
• Perform in-depth analysis and optimization to ensure the best performance on GPU architecture systems (in particular Grace/ARM-based systems). This includes support in optimization of distributed training pipelines.
• Partner with Engineering, Product, and Sales teams to develop and plan the most suitable solutions for customers.
• Enable development and growth of product features through customer feedback and proof-of-concept evaluations.

What We Need to See

• Excellent verbal, written communication, and technical presentation skills in English.
• MS/PhD or equivalent experience in Computer Science, Data Science, Electrical/Computer Engineering, Physics, Mathematics, or other Engineering fields.
• 5+ years of work or research experience with Python / C++ / other software development.
• Work experience and knowledge of modern NLP, including a good understanding of transformer, state space, diffusion, and MOE model architectures. This can include either expertise in training or optimization/compression/operation of DNNs.
• Understanding of key libraries used for NLP/LLM training (such as Megatron-LN, NeMo, DeepSpeed, etc.) and/or deployment (e.g., TensorRT-LLM, vLLM, Triton Inference Server).
• Track record in neural network performance optimization and/or training robustness.
• Person excited to work with multiple levels and teams across organizations (Engineering, Product, Sales, and Marketing teams).
• Capable of working in a constantly evolving environment without losing focus.
• Self-starter with a demeanor for growth, passion for continuous learning, and sharing findings across the team.

Ways to Stand Out from The Crowd

• Ability to conduct LLM post-training, in particular knowledge of large-scale RL.
• Track record in running large-scale training/HPC jobs with a focus on training robustness / failure resilience.
• Understanding of HPC systems: data center design, high-speed interconnect InfiniBand, Cluster Storage.

Employment Type

Full time

Location Type

Information not provided

Salary

Information not provided