Senior AI Researcher - World Foundation Models

Description

The company has spent more than 25 years advancing computer graphics, PC gaming and accelerated computing, and now applies AI to the next phase of computing, where its GPUs serve as the processing core for computers, robots and self-driving cars that understand their surroundings.

The team builds world foundation models that generate high-fidelity, temporally stable video and worlds for Physical AI, simulation and interactive uses. The role sits between applied research and production: you will validate model improvements, then turn them into production-grade checkpoints and recipes other teams can build on. The focus is human appearance, motion and action understanding. Progress is tracked through disciplined experiments, diagnostics and repeatable side-by-side evaluations, with close work alongside data, platform and product engineers so gains reach real performance and quality.

Responsibilities

• Research, implement and validate changes to model architecture and algorithms that raise video generation fidelity, with particular attention to people.
• Prototype improvements in spatial multimodal modeling, modality alignment, flow-based or diffusion-based video generation, and neural-rendering-inspired representations to boost controllability and consistency over long horizons.
• Make training and inference faster and leaner through architectural and post-training methods such as compute and memory optimization, distillation, pruning and compression.
• Define training objectives that carry models across sim-to-real and real-to-sim gaps, especially for human motion, contact and interaction across real and synthetic data.
• Build detailed, domain-specific benchmarks for world foundation models, covering both generation and understanding of video, simulation and physical environments.
• Turn research results into robust training code, production-grade checkpoints, model integrations and demos that show clear capability gains to other teams.

Requirements

• A PhD in computer science, graphics, computer engineering or a closely related field, or equivalent experience.
• 8+ years of applied research or industry experience in vision, graphics or adjacent ML areas, or similar.
• 3+ years designing, training and evaluating generative models for image, video or audio, with solid grounding in modern deep learning.
• Hands-on work improving generative models for perceptual quality and temporal stability, especially for human generation.
• Advanced skill in Python, PyTorch, C++ and CUDA, following sound research engineering practice: reproducibility, testing, profiling and experiment tracking.
• Experience training and debugging large models across multiple GPUs and/or nodes, including distributed training workflows.
• Practical knowledge of inference and runtime bottlenecks and how to optimize them.
• A sharp eye for quality, with the ability to diagnose visual artifacts such as sharpness, texture detail and temporal stability using perceptual metrics, human preference signals or learned evaluators.
• Nice to have: a track record in related research, including publications at top conferences such as NeurIPS, CVPR or ICLR, with clear evidence of improved model quality or robustness.
• Nice to have: experience using agentic workflows and AI coding assistants to speed up research and production work, such as code generation, debugging, test writing, experiment automation, benchmark development, documentation and navigating large codebases.

Conditions

Base salary depends on location, experience and other objective job-related factors. Level 4 ranges from 184,000 USD to 287,500 USD; Level 5 ranges from 224,000 USD to 356,500 USD.

Applications are accepted at least until November 3, 2026.

This is an existing vacancy.

The company uses AI tools in its recruiting process.

The company is an equal opportunity employer and does not discriminate on the basis of any characteristic protected by law.

Benefits

• Competitive salary and a comprehensive benefits package, with details for you and your family at the company's site.