Research Engineer / Research Scientist, Health

OpenAI•San Francisco, CA

1d•$295,000 - $555,000•Hybrid

About The Position

The Health team, within OpenAI’s broader Personal AGI organization, has a mission to ensure AGI improves health for all humanity. Improving human health will be one of the defining impacts of AGI. Hundreds of millions of people already turn to ChatGPT for questions about their health and millions of clinicians use it weekly to support care delivery. Increasingly capable models create an opportunity to make high-quality medical intelligence more accessible across patients and clinicians—raising the floor of human health—and accelerate the new capabilities and scientific advances that raise the ceiling of human health. Our job is to make those benefits real. We work across the full model stack—pretraining, midtraining, reinforcement learning, post-training, evaluations, harnessing, and deployment—and connect that research to the patients, clinicians, and real-world outcomes we aim to improve. We’re looking for an exceptional, hands-on researcher who wants to build frontier health capabilities and turn them into impact at scale. This is a role for someone who can take an important, underdefined problem from 0→1: identify the right bet, build what’s needed to test it, and drive it all the way to a measurable improvement in the models and products we actually ship. We’re especially excited about two kinds of people: researchers with the technical depth to move the frontier in pretraining, reinforcement learning (RL) / post-training, or evals; and researchers with real depth in developing frontier biomedical AI capabilities. Prior experience in healthcare is helpful but not required. Research excellence, velocity, ownership, and alignment with the mission are most important to us. 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.

Requirements

Have an unusually strong track record in machine learning or AI research, with the depth to push at least one important area forward: pretraining, reinforcement learning / post-training, evals (not necessarily specific to health); or substantial depth in a relevant frontier health / biomedical AI problem.
Are a hands-on builder: you can design experiments, write and debug code, work directly with data and training infrastructure, and move quickly from an idea to evidence.
Have strong research taste and can distinguish an important problem or a promising approach from an interesting one.
Own ambiguous problems end to end, and are willing to learn whatever is missing to get the job done.
Stay goal-oriented rather than method-oriented, and don’t shy away from unglamorous work if it is the highest-leverage path to impact.
Care about whether a capability will scale, generalize, and make it into the real world—not just whether it can be shown in a paper or prototype.
Enjoy working with low-ego, ambitious people across research, engineering, medicine, and product.

Nice To Haves

Prior experience in healthcare is helpful but not required.
Experience with health AI, longitudinal health data, or clinical research.

Responsibilities

Own a high-leverage research direction end to end—from deciding which problem matters and how to measure it, to designing and running experiments, to integrating successful work into frontier models.
Develop scalable methods across pretraining, midtraining, reinforcement learning, and post-training to improve health reasoning, knowledge, reliability, calibration, and behavior.
Build and study RL environments grounded in real health problems; understand which data and training setups produce robust, generalizable improvements rather than just higher benchmark scores.
Create meaningful, trustworthy, and unsaturated evaluations that tell us whether our models are actually getting better at what matters for patients and clinicians.
Advance new 0→1 capabilities in health that may have an increasing impact with upcoming increases in model intelligence and test-time compute.
Develop models and agents that can work over longitudinal, multimodal health data to produce useful predictions, surface new scientific insights, and support better individual- and population-level decisions.
Work closely with researchers, engineers, clinicians, and product teams to bring health capabilities into the models and products used by hundreds of millions of people.
Help establish that AI improves real health outcomes, not just that it can produce impressive answers or demos.