Agent Post-Training, Artifacts Research

Details

The Agent Post-Training team creates frontier agents for Codex, ChatGPT, the API, and other products. The team develops training data, environments, graders, training methods, and feedback loops for capabilities including coding, tool use, computer use, multi-agent coordination, long-horizon execution, factuality, instruction following, calibrated reasoning, and taste.

As a member of Agent Post-Training, Artifacts, you will train frontier models to create polished, useful work products, including documents, spreadsheets, slide decks, dashboards, reports, analyses, and other interactive or editable artifacts. You will help models move from vague user goals to finished artifacts with strong structure, visual taste, domain judgment, correctness, and low latency. The role involves improving the post-training stack, including reinforcement learning, data pipelines, graders, reward signals, evaluations, and behavioral analysis.

You will collaborate with researchers, engineers, product teams, infrastructure teams, and safety and alignment partners to define major model runs, measure results, and ship improvements into products used by real people.

Responsibilities

• Design and run experiments that improve agentic model behavior for complex software and plugins.
• Own end-to-end improvements to the post-training stack, including reinforcement learning, data pipelines, graders, reward signals, evaluations, diagnostics, and model-behavior analysis.
• Build evaluations and environments that expose model failures, then turn those failures into training data, product fixes, or new research directions.
• Partner with Codex and ChatGPT product teams to understand user needs and translate product signals into model improvements.
• Work on early-training and alignment interventions, including data mixtures, objectives, synthetic data, and evaluation loops that shape downstream agent behavior.
• Help determine which integrations, capabilities, and fixes are ready for inclusion in major model runs.
• Improve large-scale training and launch systems for experiment velocity, reliability, observability, reproducibility, cost, latency, and production readiness.
• Take on cross-functional projects involving model training, product infrastructure, and the production agent harness, including multi-agent systems and training against production-like environments.
• Debug difficult failures in shipped or near-shipped models and turn qualitative behavior into concrete hypotheses, experiments, and fixes.

Requirements

• Strong technical fundamentals in machine learning, software engineering, systems, statistics, or a related field.
• Hands-on experience with LLMs, reinforcement learning, RLHF/RLAIF, post-training, evaluations, graders, synthetic data, model training, coding agents, tool-using agents, or production machine learning systems.
• Ability to work on open-ended problems with noisy signals, combining research judgment with engineering execution.
• Interest in product impact and model behavior, including what makes an agent useful, reliable, honest, tasteful, and easy to work with.
• Ability to turn a vague behavioral problem into a concrete experiment by defining a hypothesis, building a pipeline, running a model, analyzing results, and determining next steps.
• Ability to collaborate across research, product, infrastructure, data, evaluations, and safety functions.
• Willingness to build reliable systems and processes as required by the team.
• Interest in training and shipping models for developers, enterprises, researchers, and everyday users.
• Prior background in consulting, finance, marketing, operations, or data science may be helpful.

About OpenAI

OpenAI is an AI research and deployment company dedicated to ensuring that general-purpose artificial intelligence benefits all of humanity. OpenAI is an equal opportunity employer and provides reasonable accommodations to applicants with disabilities. Background checks are administered in accordance with applicable law.