Jobs

Research Engineer/Research Scientist-Post Training

OpenAI

Apply on Ashby
Location
San Francisco, California, United States
Track
AI Researcher
Salary
$380K–$500K / yr
Posted
June 25, 2026
Source
Ashby

Job description

About the Team

The Post-Training team advances OpenAI’s frontier models and prepares them for deployment in ChatGPT, Codex, the API, and future products. We work at the intersection of research and product to make models more capable, useful, safe, and reliable for millions of users.

We define the capabilities our models should have, build the training signals that teach those abilities, and run the experiments that make them real. Our work spans reinforcement learning, model behavior and alignment, reasoning, coding, and agentic capabilities such as tool use, computer use, and long-horizon execution. We build the data, environments, evaluations, and training methods that shape model behavior, then carry those improvements through major training runs and into the products people use.

About the Role

As a Research Engineer / Research Scientist on Post-Training, you will research and develop improvements to the capabilities, reliability, and real-world performance of OpenAI’s models. You might own a research direction, develop new training methods, build infrastructure for large-scale experiments, or create evaluations that reveal where models fail. You will take ideas from early experimentation through integration and deployment.

We’re looking for people with strong machine learning engineering skills and research experience who are excited about product-driven research. This role is intentionally broad: you should be comfortable taking an ambiguous problem and making progress across research, engineering, data, and evaluations. You will collaborate closely with researchers, product and infrastructure teams, and safety and alignment partners to translate user needs into model improvements and help determine what is ready for major model runs.

This role is based in San Francisco, CA. We use a hybrid work model of 3 days in the office per week and offer relocation assistance to new employees.

In this role, you might

  • Own and pursue a research agenda to improve model capabilities, behavior, and performance, from defining hypotheses and running experiments to analyzing results and shipping improvements.
  • Develop and improve post-training methods and systems, including reinforcement learning, data mixtures, synthetic data, reward signals, graders, and training pipelines.
  • Build robust evaluations and environments that measure progress, expose model failures, and turn those findings into better training data, methods, and product experiences.
  • Design, implement, test, and debug code across the research stack, improving the reliability, efficiency, observability, and reproducibility of large-scale experiments.
  • Collaborate with research and product teams across ChatGPT, Codex, and the API to understand user needs and translate real-world feedback into model improvements.
  • Advance capabilities such as reasoning, factuality, instruction following, coding, tool use, and long-horizon task execution, with opportunities to work on computer use and multi-agent collaboration.
  • Partner with safety and alignment teams to develop useful, reliable model behavior and evaluate readiness for deployment.

You might thrive in this role if you

  • Have strong fundamentals in machine learning and software engineering, with research or applied experience developing and improving machine learning models.
  • Have hands-on experience in one or more relevant areas, such as large language models, reinforcement learning, RLHF/RLAIF, post-training, evaluations, synthetic data, agentic systems, or production ML systems.
  • Can turn an ambiguous model-behavior problem into a concrete experiment: define the hypothesis, build the pipeline, run the model, analyze the results, and decide what to do next.
  • Are comfortable navigating and debugging a large ML codebase and building dependable systems that support research at scale.
  • Care about product impact and how models behave in practice, including what makes them useful, reliable, honest, and easy to work with.
  • Thrive in a dynamic, technically complex environment where the path is unclear and progress requires both research judgment and engineering execution.
  • Communicate clearly and collaborate across research, product, infrastructure, data, evaluations, and safety teams.

Similar roles

Get roles like this in your inbox

New agentic AI jobs, curated every Thursday. No spam.

Apply on Ashby