An Agentic AI Framework to Accelerate Scientific Discovery in Plant Phenotyping

2026-06-30

AI Summary

The explosive growth of high-throughput plant phenotyping data has far outpaced manual analysis capabilities, creating a bottleneck in scientific discovery. This work proposes the first dual-agent, end-to-end AI framework that transforms phenotyping platforms into interactive, autonomous discovery systems through human–AI collaboration. The system employs a natural language–driven workflow wherein a conversational Co-Scientist agent and a headless Compute agent operate in isolated security and resource domains. Integrated with a Vision Transformer–based segmentation model and the Frontier exascale supercomputer, the framework enables federated computation, controlled data movement, and full provenance tracking. By overcoming limitations of conventional cloud-native agents, it compresses analysis pipelines—typically requiring days to weeks—into sub-second interactive loops, thereby supporting real-time inference, recommendation, and responsive query resolution.

Abstract

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.

Problem

• plant phenotyping
• high-throughput imaging
• data analysis bottleneck
• scientific discovery
• trait extraction

Innovation

• Agentic AI
• Plant Phenotyping
• Vision Transformer
• Exascale Computing
• Scientific Discovery Automation