Senior Multimodal RL Scientist — Agentic AI Systems

NVIDIA; Myrtle Point (OR); On-site; USD 184,000 - 288,000; Full time

Benefits offered by this job

Equity

Benefits package

Job summary

NVIDIA in the United States is seeking researchers to develop, train, and deploy large language models for agentic systems that combine audio-visual reasoning, tool usage, and document understanding.

You will advance post-training methods, such as instruction tuning, preference optimization, and RLHF/RLVR/MOPD, to improve multimodal agents for complex tasks across digital and physical environments.

Qualifications

• Master's degree (or PhD) in CS/AI or applied math with 8+ years of experience.
• Strong Python and PyTorch for scalable model development.
• Deep learning fundamentals, Transformers, and MoE architectures.
• Foundational RL algorithms and reward design.
• Experience with post-training multimodal models and multimodal reasoning.

Responsibilities

• Apply research to develop, train, fine-tune, and deploy large language models for agentic systems across audio-visual reasoning, tool usage, and document understanding.
• Advance post-training alignment methods including instruction tuning and RLHF/RLVR/MOPD to improve multimodal agents.
• Lead collection, development, and benchmarking of multimodal datasets; ensure high-quality evaluation of accuracy, safety, and task completion.
• Manage model development life cycles, including dataset versioning, experiment tracking, and evaluation pipelines.

Skills

Python

PyTorch

Transformers

MoE models

Reinforcement learning

Experiment tracking

Dataset versioning

Multimodal reasoning

Education

Master's/PhD in CS/AI/Applied Math

Job description

NVIDIA in the United States is seeking researchers to develop, train, and deploy large language models for agentic systems that combine audio-visual reasoning, tool usage, and document understanding.

You will advance post-training methods, such as instruction tuning, preference optimization, and RLHF/RLVR/MOPD, to improve multimodal agents for complex tasks across digital and physical environments.