DDAP: A Controlled Agentic Framework for Non-AI Expert Scientists
The Domain-Driven Adaptable AI Pipelines (DDAP) is an innovative framework designed to assist researchers without AI expertise in constructing AI pipelines. Detailed in arXiv:2605.18764, DDAP operates as a controlled, agentic system with human oversight, utilizing large language models to navigate users through four phases: defining the problem, specifying the compute environment, generating the pipeline, and producing code. This framework focuses on areas such as Medical Sciences, Agriculture, and Social Sciences, tackling the difficulty posed by the necessary expertise for comprehensive AI system creation.
Key facts
- DDAP stands for Domain-Driven Adaptable AI Pipelines.
- It is a controlled, human-in-the-loop, agentic framework.
- It leverages large language models to guide users.
- The development process has four stages: problem definition, compute environment specification, pipeline generation, and code generation.
- It is designed for non-AI expert scientists.
- It supports fields such as Medical Sciences, Agriculture, and Social Sciences.
- The framework adapts to domain context and user expertise.
- The paper is available on arXiv with ID 2605.18764.
Entities
Institutions
- arXiv