OpenAI26.06.2026
Agent Post-Training, Computer Use Research
Зарплата не указана
Полная занятостьSan Francisco
Обязанности
- 01Design and run experiments that improve agentic model behavior for complex computer use
- 02Own end-to-end improvements to the post-training stack, including RL, data pipelines, graders, reward signals, evals, diagnostics, and model-behavior analysis
- 03Build evals and environments that expose the next set of model failures, then turn those failures into training data, product fixes, or new research directions
- 04Partner with Codex and ChatGPT product teams to understand what users need and translate product signal into model improvements
- 05Work on early-training and alignment interventions, including data mixtures, objectives, synthetic data, and eval loops that shape downstream agent behavior
- 06Help decide which integrations, capabilities, and fixes are ready for inclusion in major model runs
- 07Improve the machinery for large-scale training and launch: experiment velocity, reliability, observability, reproducibility, cost, latency, and production readiness
- 08Take on cross-functional projects that touch model training, product infrastructure, and the production agent harness
- 09Debug hard failures in shipped or near-shipped models and turn messy qualitative behavior into concrete hypotheses, experiments, and fixes
Требования
- 01Strong technical fundamentals in machine learning, software engineering, systems, statistics, or a related field
- 02Hands-on experience with LLMs, RL, RLHF/RLAIF, post-training, evals, graders, synthetic data, model training, coding agents, tool-using agents, or production ML systems
- 03Ability to move from a vague behavioral problem to a concrete experiment: define the hypothesis, build the pipeline, run the model, analyze the result, and decide what to do next
- 04Comfortable working across research, product, infrastructure, data, evals, and safety boundaries
- 05Ability to communicate clearly with different groups
- 06Willingness to build load-bearing systems and processes when needed