AWS AI: Agentic AI

AWS AI aims to advance agentic AI research by funding development of open-source tools and research that benefit the AI community at large, or impactful research related to agents.

We welcome proposals related to agentic AI in the areas below:

1. Retrieval across structured and unstructured data
2. Tool/API calling over long horizons
3. Human interfaces, such as computer use, when programmatic APIs don’t exist
4. Self analyzing/reflecting, catching mistakes, backtracking, new path finding capabilities
5. Reward models, programmatic and learned verifiers
6. Automated reasoning and neuro-symbolic techniques for output synthesis and verification
7. Safety and responsible AI for agentic behavior
8. Automated feedback loops for agents to improve
9. Multi-agent collaborations
10. Protocols for interfacing with data sources, tools/APIs, and other agents
11. Application domains, including but not limited to, software engineering, enterprise, healthcare & life sciences, finance, media & entertainment, consumer agentic assistants

Up to $120,000 USD