Research Engineer Scientist Post Training
Letta is a San Francisco AI company building persistent, stateful agents that learn from experience through memory, identity, and evolving capabilities.
Funding history
About Letta
Letta develops and operates an AI agent platform, including Letta Agent and open-source Letta Code, for building, deploying, and using stateful agents across local and cloud environments.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will train models for agentic tool use and context management. You will develop continuous post-deployment model updates, run experiments on data mixtures and training algorithms, build synthetic-data infrastructure, create agentic-capability evaluations, and publish research through papers, reports, blog posts, and open-source code.
Requirements
- Python
- Deep learning framework
- PyTorch
- Post-training
- Supervised fine-tuning
- Reinforcement learning
- Reward model
- Preference learning
- Empirical rigor
- Impactful research
- Publications
- Open-source contributions
- Real-world impact
Responsibilities
- Train models for agentic tool use and context management
- Design mechanisms for continuous post-deployment model-weight updates without catastrophic forgetting
- Design and run experiments on data mixtures, training algorithms, and models
- Build infrastructure to generate and collect synthetic data at scale
- Build evaluations for agentic capabilities
- Publish research through papers, technical reports, blog posts, and open-source code
Hiring Process
Initial screen (30 min) → technical screen (1–1.5 hours) → paid in-person work trial (2 days onsite in San Francisco).