Post-Training Agent: Mastering Tool Integrations

OpenAI; San Francisco (CA); On-site; USD 380,000 - 500,000; Full time

Job summary

OpenAI is seeking an Agent Post-Training, Connectors Research member in San Francisco to teach models how to interface with professional software using code. You’ll train agents to use code, APIs, tools, and structured integrations to operate Slack, Google Workspace, GitHub, Notion, and more, enabling actions across a user’s digital context.

You’ll collaborate with researchers, engineers, product teams, and safety partners to decide what goes into major model runs, measure outcomes, and ship

Qualifications

• Strong technical fundamentals in ML, software engineering, systems, statistics, or related field.
• Hands-on experience with LLMs, RL, RLHF/RLAIF, post-training, evals, graders, synthetic data, model training, coding agents, tool-using agents, or production ML systems.
• Excited by open-ended problems; signal is noisy; requires research taste and engineering execution.
• Care about product impact and model behavior, not just benchmark movement.
• Capable of turning vague problems into concrete experiments: hypothesize, pipeline, run, analyze, decide.
• Comfortable working across research, product, infra, data, evals, and safety.

Responsibilities

• Design and run experiments to improve agentic model behavior for complex software and plugins.
• Own end-to-end improvements to the post-training stack: RL, data pipelines, graders, reward signals, evals, diagnostics, and model-behavior analysis.
• Build evals and environments to expose next model failures and turn them into training data or product fixes.
• Collaborate with Codex and ChatGPT product teams to translate user needs into model improvements.
• Work on early-training and alignment interventions, including data mixtures and synthetic data.
• Decide which integrations and fixes are ready for major model runs.
• Improve machinery for large-scale training, launch velocity, observability, and production readiness.
• Tackle cross-functional projects across model training, product infra, and the production harness.
• Debug hard failures in shipped models and convert qualitative behavior into hypotheses and fixes.

Job description

OpenAI is seeking an Agent Post-Training, Connectors Research member in San Francisco to teach models how to interface with professional software using code. You’ll train agents to use code, APIs, tools, and structured integrations to operate Slack, Google Workspace, GitHub, Notion, and more, enabling actions across a user’s digital context.

You’ll collaborate with researchers, engineers, product teams, and safety partners to decide what goes into major model runs, measure outcomes, and ship