JOB-1833
An AI startup building software agents that perform engineering work in the physical world is hiring a research scientist to lead how those agents keep learning after deployment. This is an individual contributor position at principal level, and the results go directly to customers rather than staying in a lab.
Your focus
- Define from scratch the research agenda covering memory formation, self-improvement, rapid adaptation and continual learning
- Steer the technical approach at principal level within a small group of researchers and engineers
- Take promising ideas from the roadmap, sharpen them and start experiments right away
- Carry research into production alongside AI and software engineers, covering data creation, evaluation and folding results into the product
- Chase breakthroughs that make the agents more capable, while publishing strong results
About you
- A doctorate or comparable experience in computer science, robotics, engineering, mathematics, physics or a neighboring discipline
- Roughly three to seven years of research that has been put to use in practice, with publications at leading AI and ML venues
- Deep background spanning meta-learning, lifelong learning, systems that improve themselves, memory mechanisms and adapting models; agentic work over long horizons is a major advantage
- Practical experience with post-training, including supervised fine-tuning and reinforcement learning
- Strong command of Python, JAX or PyTorch and large-scale distributed training
- Motivated by a scrappy early-stage setting where you own outcomes
On-site in the San Francisco Bay Area. Not able to offer visa sponsorship or transfers.
