Agent Post-Training, Personality

Details

The Agent Post-Training team creates the frontier agents OpenAI ships to the world. The team trains models used in Codex, ChatGPT, the API, and other frontier products to produce persistent, proactive intelligence that can operate computers, collaborate with people and other agents, and expand what people and organizations can imagine, attempt, and achieve. The team builds the data, environments, graders, training methods, and feedback loops that shape what OpenAI's agents can do and what they are like to work with, and carries those improvements through major training runs into products used daily.

Responsibilities

• Define what next-generation agents should be able to do and build the training signal that teaches those abilities.
• Develop a rigorous understanding of what makes an agent a great collaborator across professional, creative, technical, and everyday work.
• Turn qualitative judgments about model behavior into concrete hypotheses, evaluations, graders, and training interventions.
• Study explicit and implicit user signals to understand which behaviors create trust, satisfaction, continued use, and successful outcomes.
• Work with human experts and trainers to produce high-quality rollouts and preference data that capture excellent collaborative behavior.
• Improve reward models and reinforcement learning (RL) objectives for model behaviors.
• Work with pretraining and early-training teams on data mixtures, objectives, synthetic data, and other upstream choices that shape downstream personality.
• Build sustainable pipelines for updating older training data as understanding of model behavior evolves.
• Partner closely with product teams (ChatGPT, Codex, etc.) to turn consumer insight into model improvements and validate them in real workflows.
• Own projects end to end: observe behavioral failures through experimentation, training, evaluation, and launch.

Requirements

• Strong technical foundations in one or more of: machine learning, software engineering, statistics, behavioral science, HCI, or a related field.
• Ability to translate subjective product questions into falsifiable hypotheses and rigorous evaluations while preserving nuance.
• Strong taste for model behavior: able to analyze user feedback and explain why responses feel thoughtful, natural, and useful (or not).
• Experience with LLMs, post-training, RL/RLHF, reward modeling, evaluations, synthetic data, pretraining data, or production ML systems.
• Comfortable working with researchers, engineers, product teams, designers, domain experts, human-data teams, and safety boundaries; clear communicator across these groups.
• Comfortable with ambiguous capability problems where signals are noisy and solutions may involve data, training, evals, product changes, or combinations thereof.
• Willingness to build sustainable systems and processes (data pipelines, graders, training methods) needed for productionizing improvements.

Benefits

• Base salary range: $295,000 - $445,000 (base pay may vary by location, skills, and experience). Total compensation may include equity and performance-related bonuses.
• Medical, dental, and vision insurance with employer contributions to Health Savings Accounts.
• Pre-tax accounts (Health FSA, Dependent Care FSA, commuter expenses).
• 401(k) with employer match.
• Paid parental leave, paid medical and caregiver leave, flexible PTO and paid company holidays/closures.
• Mental health and wellness support; employer-paid basic life and disability coverage.
• Annual learning and development stipend.
• Daily meals in offices and meal delivery credits as eligible.
• Relocation support for eligible employees; additional fringe benefits may be provided.

About OpenAI

OpenAI is an AI research and deployment company dedicated to ensuring that general-purpose artificial intelligence benefits all of humanity. The company emphasizes safety, inclusion of diverse perspectives, and provides reasonable accommodations to applicants with disabilities. Background checks are administered in accordance with applicable law.