Kinetic Systems is an early-stage startup working at the intersection of computer-use agents, human data, and healthcare. Our mission is to advance the capabilities of frontier AI models on economically meaningful healthcare tasks by building novel datasets, environments, and models. We were founded in 2025 out of the Stanford PhD program and backed by Tier 1 VCs.
The RoleAs an Applied AI Intern, you’ll help drive our research agenda towards advancing AI model capabilities for real-world healthcare tasks. You'll build novel evals and benchmarks to identify capability gaps in current models, publish academic papers, and post-train models to push the SOTA while staying grounded to the demonstrated needs of our healthcare partners.
What You’ll DoDevelop novel evals, RL environments, and benchmarks to reflect real-world healthcare workflows
Develop, train, and evaluate computer-use agents for complex healthcare interfaces
Build and maintain data pipelines to transform raw human data into high-quality training and evaluation assets
Write and publish papers in academic conferences
Work across the stack: models, tooling, infra, product, and internal workflows
Published at 1+ first-author papers in a top ML conference (NeurIPS, ICLR, ICML, etc.)
Trained a 1B+ param model from scratch
Come from a research background (preference for MS or PhD) in at least one of these fields: Computer-Use Agents, Vision-Language Models, Computer Vision, Robotics, RL
Have significant experience with PyTorch, HuggingFace, or similar libraries
Are high-agency and comfortable owning large, ambiguous problem spaces
Are comfortable working long hours in a high-intensity, early-stage environment
Are excited to be on-site in SF and collaborate closely with a small team
Are interested in healthcare as an application (prior background not necessary)
You must be in-person in SF for the duration of the internship. We will provide support to help relocate.
You must be authorized to work in the US (we support eVerify)
You can start within the next month
Skills Required
- 1+ first-author publication in a top ML conference (NeurIPS, ICLR, ICML, etc.)
- Trained a 1B+ parameter model from scratch
- Research background in Computer-Use Agents, Vision-Language Models, Computer Vision, Robotics, or Reinforcement Learning
- MS or PhD preferred
- Significant experience with PyTorch, HuggingFace, or similar libraries
- High-agency and ability to own large, ambiguous problem spaces
- Comfortable working long hours in a high-intensity, early-stage environment
- Willingness and ability to be on-site in San Francisco for the duration of the internship (relocation support provided)
- Authorized to work in the US (eVerify)
- Available to start within the next month
- Interest in healthcare applications (prior healthcare background not required)
What We Do
Kinetic Systems is an applied AI research lab focused on improving frontier models to solve economically meaningful healthcare tasks. Founded in 2025 out of the Stanford PhD program and backed by Tier 1 VCs, its mission is to advance the capabilities of frontier AI models by building novel datasets, environments, and models. The company operates at the intersection of computer-use agents, human data, and healthcare.
