Multimodal AI Research Scientist – Personalization & RLHF

OpenAI; California (MO); Hybrid; USD 180,000 - 240,000; Full time

Benefits offered by this job

Hybrid work model

Relocation assistance

Job summary

OpenAI is seeking a Research Engineer/Scientist for the Future of Computing Research team in San Francisco, CA, to advance RLHF and post-training for personalized, multimodal AI systems. The role emphasizes learning foundations, evaluation, and long-term user value in real-world settings.

You will prototype training recipes, reward formulations, data pipelines, and evaluation suites while collaborating with safety researchers to ensure trustworthy, interpretable systems that scale across

Qualifications

• Experience in RLHF, reward modeling, or post-training for large models.
• Experience in reinforcement learning, ranking, recommender systems, personalization, memory, or human-in-the-loop evaluation.
• Ability to design clean experiments and reliable evals.

Responsibilities

• Develop RLHF and post-training methods for multimodal models.
• Build reward models and preference-learning pipelines for adaptive, personalized model behavior.
• Design datasets, rubrics, and evaluation frameworks that capture user preferences and long-term value.
• Run experiments on policy improvement using explicit/implicit feedback and model-based grading.
• Collaborate with safety researchers to ensure alignment and bounded adaptation.

Skills

RLHF

Reward modeling

Preference optimization

Post-training for large models

Reinforcement learning

Ranking

Personalization

Memory

Human-in-the-loop evaluation

Multimodal AI

Job description

OpenAI is seeking a Research Engineer/Scientist for the Future of Computing Research team in San Francisco, CA, to advance RLHF and post-training for personalized, multimodal AI systems. The role emphasizes learning foundations, evaluation, and long-term user value in real-world settings.

You will prototype training recipes, reward formulations, data pipelines, and evaluation suites while collaborating with safety researchers to ensure trustworthy, interpretable systems that scale across