Senior Solutions Architect, Robotics Foundation Model Training

This is a fully remote position, open to applicants in California.

📋 Description

• Collaborate with researchers and ML engineers to design and enhance comprehensive training workflows for robotics foundation models, such as World Models, VLAs, and WAMs.

• Develop proof-of-concepts, reference architectures, and agentic workflows to expedite experimentation, benchmarking, and model enhancement of NVIDIA’s robotics open model platforms, including Cosmos and GR00T.

• Scale pre-training, fine-tuning, and reinforcement learning tasks across multi-GPU and multi-node configurations.

• Enhance utilization, throughput, and memory efficiency.

• Detect and resolve data pipeline bottlenecks in storage, networking, preprocessing, and data loading for multimodal datasets, including video, sensor data, and trajectories.

• Partner with NVIDIA product and engineering teams to offer insights that shape the future of Physical AI platforms.

• Engage with research, engineering, and customer teams to influence product direction and the adoption of applied AI.

⛳️ Requirements

• MS, PhD, or equivalent experience in Computer Science, Artificial Intelligence, Electrical or Computer Engineering, Robotics, or a related discipline.

• Over 5 years of industry or research experience in deep learning, distributed computing, or large-scale model training.

• Practical experience in training or optimizing multimodal or foundation models, preferably in robotics environments.

• Familiarity with the AI model lifecycle, encompassing pre-training, supervised fine-tuning, reinforcement learning or other post-training methods, evaluation, and model optimization.

• Strong knowledge of distributed training methodologies, including data/model/pipeline parallelism, sharding, and checkpointing, on multi-GPU or multi-node setups.

• Proficiency with multimodal training frameworks such as PyTorch, NVIDIA NeMo, JAX, or Hugging Face Transformers.

• Experience in constructing or utilizing high-throughput data pipelines for large-scale training, focusing on storage bandwidth, network throughput, and preprocessing techniques like decoding, tokenization, and batching.

• Excellent communication skills with the capability to effectively collaborate with Researchers, Engineers, and executives.

• Acquaintance with NVIDIA AI and robotics platforms such as Cosmos, GR00T, NeMo, Isaac Sim, and Isaac Lab.

• Background in robotics AI tasks, including reinforcement learning in simulation and synthetic data generation.

• Experience profiling and optimizing workloads using Nsight Systems, Nsight Compute, or PyTorch Profiler.

• Proven track record of enhancing training efficiency and scaling performance.

• Experience in developing agentic workflows for automated experimentation, model evaluation, data analysis, or research acceleration.

🏝️ Benefits

• Equity

• Benefits