Research Engineer / Scientist, Self-Improvement
Job Description
Letta, founded by the creators of MemGPT from UC Berkeley's Sky Computing Lab, is at the forefront of developing self-improving artificial intelligence. The company focuses on creating agents that continually learn from experience and adapt over time, already powering production systems at companies like 11x and BILT Rewards.
As a Research Engineer/Scientist in Self-Improvement, you'll be tackling AI's hardest problems, working on foundational methods that enable AI systems to learn and evolve through experience. The role involves developing novel memory systems, context management techniques, and agent architectures that push the boundaries of AI capabilities towards human-like learning.
The position is based in downtown San Francisco, requiring full-time in-person presence 5 days a week. You'll be joining a world-class, tight-knit team of AI researchers and engineers, working in a flat organizational structure where everyone contributes as an individual contributor.
Key research areas include Memory & Learning Systems, Agent Architectures, and Open Research & Impact. The ideal candidate will have deep expertise in machine learning, particularly in LLMs and continual learning, with a proven track record of impactful research. The role requires someone who thrives in ambiguity, can drive their own agenda, and is willing to work beyond traditional hours to achieve breakthrough results.
This is an opportunity to be part of a potentially transformative company in its early stages, similar to being at OpenAI when it was just starting. The company values direct customer engagement, cross-stack work, and maintains a strong stance against closed frontier AI controlled by private tech companies. The interview process includes an initial screen, technical screen, and a paid in-person work trial in San Francisco.
Responsibilities For Research Engineer / Scientist, Self-Improvement
- Develop foundational methods for AI systems to learn and self-improve over time
- Design novel memory systems for agent learning and improvement
- Develop context management techniques
- Research multi-agent coordination and knowledge sharing
- Develop evaluation frameworks for measuring learning and adaptation
- Pioneer new approaches to agent reasoning and planning
- Publish research through papers, technical reports, and blog posts
- Collaborate with research community on self-improvement and continual learning
Requirements For Research Engineer / Scientist, Self-Improvement
- Deep expertise in machine learning, particularly in LLMs, continual learning, and/or LLM agent architectures
- Track record of impactful research (breakthrough publications and/or open source contributions)
- Desire for real-world impact beyond pure academic work
- Willingness to work beyond 9-5 schedule
- Ability to work across the stack and engage directly with customers
- Ability to thrive in ambiguity and drive own agenda
