Research Scientist - Multimodal Agent, Consumer Devices

OpenAI; United States; Remote; USD 150,000 - 230,000; Full time

Job summary

OpenAI seeks a Research Scientist to develop multimodal AI systems for personalized user experiences. You will advance reinforcement learning from human feedback and preference-learning methods, and you will design datasets and evaluation frameworks that prioritize user preferences and contextual appropriateness to ensure models adapt effectively over time.

Collaborating with safety researchers, you will maintain alignment and interpretability in personalized AI systems, prototyping training

Qualifications

• Candidates with a strong machine learning background and experience in reinforcement learning and evaluation methodologies should apply.
• Focusing on advancing multimodal AI systems by developing RLHF and preference-learning methods.
• Designing datasets and evaluation frameworks that prioritize user preferences and contextual appropriateness.
• Collaborating with safety researchers to maintain alignment and interpretability in personalized AI systems.
• Prototyping training recipes and evaluation suites to shape product-relevant behaviors while addressing long-term value and trustworthiness in AI interactions.

Responsibilities

• Develop multimodal AI systems using RLHF and preference-learning methods.
• Design and curate datasets and evaluation frameworks prioritizing user preferences.
• Prototype training recipes and evaluation suites that steer product-relevant behaviors.

Skills

Machine Learning

Reinforcement Learning

Evaluation Methodologies

Multimodal AI

RLHF

Dataset Design

Collaboration

Tools

PyTorch

TensorFlow

RL Frameworks

Job description

OpenAI seeks a Research Scientist to develop multimodal AI systems for personalized user experiences. Candidates with a strong machine learning background and experience in reinforcement learning and evaluation methodologies should apply.

In this role, you will focus on advancing multimodal AI systems by developing reinforcement learning from human feedback (RLHF) and preference-learning methods. Your work will involve designing datasets and evaluation frameworks that prioritize user preferences and contextual appropriateness, ensuring models adapt effectively over time.

Collaboration is key, as you will work closely with safety researchers to maintain alignment and interpretability in personalized AI systems. Success in this position means effectively prototyping training recipes and evaluation suites that shape product-relevant behaviors while addressing the long-term value and trustworthiness of AI interactions.