Senior Deep Learning Scientist, Multimodal Agentic Reinforcement Learning

Quick Facts

• Role: Senior Deep Learning Scientist (Multimodal Agentic Reinforcement Learning)

Description

You will apply foundational and applied research to develop, train, fine-tune, and deploy large language models for agentic systems spanning audio-visual reasoning, tool usage, and document understanding. You will advance post-training and alignment methods (instruction tuning, preference optimization, RLHF/RLVR/MOPD) and research agentic reasoning and grounded perception for long-horizon tasks in digital and physical environments. You will also lead multimodal dataset collection, development, and benchmarking for accuracy, safety, and task-completion success.

Responsibilities

• Develop and deploy agentic multimodal LLMs for audio-visual reasoning, tools, and document understanding

• Improve multimodal agents using instruction tuning, preference optimization, and RLHF/RLVR/MOPD

• Research planning, tool execution, and grounded perception for long-horizon tasks across environments

• Collect, build, and benchmark high-quality multimodal datasets with evaluation for accuracy, safety, and success

Requirements

• Master’s degree (or equivalent) or PhD in Computer Science, AI, or Applied Math

• 8+ years of relevant work experience

• Strong Python programming skills for scalable model development

• Knowledge of ML/DL techniques and foundation model architectures (Transformers, mixture-of-experts)

• Foundational reinforcement learning knowledge, including MDPs, policies, and reward design

• Hands-on experience in post-training multimodal omni-modality models and voice interaction

• Ability to manage model development lifecycle: dataset versioning, experiment tracking, evaluation pipelines

Benefits

• Highly competitive salaries

• Comprehensive benefits package and eligibility for equity and benefits

• Base salary range: 184,000 USD - 287,500 USD