NVIDIA
Senior Deep Learning Engineer, Cosmo 3D Spatial
$224K–$356K
US, CA, Santa Clara · Senior · Full-time · Posted Sep 15, 2026
- Experience
- 12+ years.
- Education
- MS or PhD in Computer Science, Electrical/Computer Engineering, Robotics, or a related field, or equivalent experience.
About the job
Own the 3D data engine, training verification, and evaluation systems that develop and validate spatial reasoning and perception in NVIDIA Cosmos models.
What you'll do
- Build and manage large-scale 3D-grounded training data, annotation, and data-quality pipelines.
- Run and verify pre-training and supervised fine-tuning; diagnose regressions and link capability changes to data and recipe decisions.
- Build spatial evaluation benchmarks and work with research scientists to test hypotheses and guide model development.
What they ask for
- 12+ years of experience building deep learning systems in Python with PyTorch or JAX on Linux.
- Deep expertise in 3D computer vision or related areas, with practical ability to produce and validate 3D ground truth at scale.
- Hands-on experience with vision-language models and building data and evaluations to improve visual grounding and reasoning.
- Experience with large-scale multimodal data pipelines and training large models on multi-GPU, multi-node clusters.
Nice to have
- PhD and/or publications at CVPR, ICCV, ECCV, NeurIPS, ICLR, or CoRL in 3D vision, multimodal learning, or embodied AI.
- Experience with web-scale or petabyte-scale video corpora, 3D foundation models, vision-language post-training, or evaluation infrastructure.
Pay
224,000 USD - 356,500 USD (base salary)
Tools
Python, PyTorch, JAX, Linux, Ray, Spark, Slurm, VLMEvalKit
AI focus
Develop and validate 3D spatial reasoning and perception in world foundation models for Physical AI, including grounding, geometry, and embodied spatial reasoning.
Model-written summary · the original posting is authoritative