Search papers, labs, and topics across Lattice.
This paper introduces an agentic AI framework designed to transform high-throughput plant phenotyping by automating the analysis of image-derived datasets, which currently outpace human capabilities. By integrating a conversational Co-Scientist Agent with a Compute Agent on an exascale supercomputer, the framework enables real-time interaction and iterative analysis, drastically reducing the time from data acquisition to insight. The system not only enhances efficiency but also ensures comprehensive provenance tracking, addressing critical challenges in data management and analysis in scientific research.
Transforming plant phenotyping from a data factory into an interactive, autonomous discovery platform cuts analysis time from weeks to mere seconds.
High-throughput plant phenotyping now generates image derived datasets far faster than scientists can analyze them. At Oak Ridge National Laboratory's Advanced Plant Phenotyping Laboratory (APPL), automated stations image hundreds of plants daily across multiple remote sensing modalities; yet, trait extraction and interpretation remain manual, expert-bound, and strictly post-hoc, making analysis, not acquisition, the binding constraint on discovery. We present an end-to-end agentic AI framework that turns the facility from a data factory into an interactive autonomous, discovery platform, where scientists partner with AI agents to accelerate time to insight. A conversational Co-Scientist Agent translates a scientist's natural-language question into a structured analysis plan, and a headless Compute Agent dispatches Vision Transformer segmentation and trait extraction on the Frontier exascale supercomputer. The two agents run in separate security and resource domains and communicate over a secure, token-authenticated streaming channel, a design that accounts for the federation, data-movement, and provenance realities cloud-native agentic frameworks ignore, ensuring end-to-end provenance is captured for every interaction. The framework turns a days- to weeks-long analysis process into an interactive loop where agents reason over results, recommend next analyses, and respond to follow-up questions in seconds.
