Senior Multimodal 3D/4D World Model Engineer

NVIDIA; Seattle (WA); On-site; USD 184,000 - 357,000; Full time

Benefits offered by this job

Equity

Benefits

Job summary

NVIDIA Seattle seeks an engineer to advance the DRIVE Road Structure, Online Mapping, and Context Fusion program. You will craft a learned 3D/4D world model that fuses navigation, ego-motion, perception, and signals to support robust L3/L4 autonomous driving updates.

You will collaborate across perception, mapping, prediction, planning, and simulation teams to transform research into production-ready world representations and to push the boundaries of multimodal fusion and scalable training.

Qualifications

• PhD/ MS/ BS in CS, robotics, electrical engineering or related field, or equivalent experience.
• 8+ years in AV/robotics with 2+ years leadership in technical area.
• Strong experience building production-quality sensor fusion or autonomous driving systems.
• Experience with learning-based multimodal perception or fusion across cameras, LiDAR, radar, map, and ego-motion signals.
• Solid understanding of 3D geometry, calibration, temporal synchronization, and uncertainty estimation.
• Experience with deep-learning methods for 3D perception, occupancy, or motion planning.
• Strong C++ and Python, PyTorch, CUDA, and GPU deployment experience.
• Experience with Transformer/VLM/multimodal architectures and scalable training.

Responsibilities

• Design and develop multimodal sensor-fusion systems that transform sensor history and context into a unified spatiotemporal world representation.
• Build architectures that fuse camera, LiDAR, radar, and vehicle-state inputs with calibration, synchronization, and uncertainty handling.
• Develop end-to-end models producing driving-relevant outputs from a shared scene representation (lanes, boundaries, traffic controls, occupancy).
• Create scalable fusion architectures (Transformer-based, BEV/voxel representations, temporal context).
• Develop training, fine-tuning, and evaluation pipelines for large-scale multimodal datasets with multi-task objectives and metrics.
• Explore foundation-model approaches for autonomous driving (vision-language, multimodal pre-training).
• Collaborate with perception, mapping, prediction, planning, and simulation teams to deploy robust AV systems.
• Build tooling for debugging model failures, cross-sensor disagreement, and distribution shift.

Skills

C++ & Python

PyTorch

Transformer models

CUDA / TensorRT

Distributed training

Autonomy knowledge

Education

BS/MS/PhD in CS/Robotics/EE

Tools

CUDA

TensorRT

Distributed tools

Job description

NVIDIA Seattle seeks an engineer to advance the DRIVE Road Structure, Online Mapping, and Context Fusion program. You will craft a learned 3D/4D world model that fuses navigation, ego-motion, perception, and signals to support robust L3/L4 autonomous driving updates.

You will collaborate across perception, mapping, prediction, planning, and simulation teams to transform research into production-ready world representations and to push the boundaries of multimodal fusion and scalable training.