Track 2 · Engineering Systems

LLM Application
in Agriculture

Encoding expert knowledge into language models for agricultural decision-making

llm-agri-reasoning ❯ model.encode(domain_knowledge) papers: 5,000 • expert rules: 128 embedding ████████████▓░░ 84% ❯ model.reason(observation) input: {EC: 2.8, moisture: 62%} → retrieving relevant knowledge... → cross-referencing 12 experiences ❯ model.act_and_learn() ✓ decision: adjust irrigation −15% ✓ memory updated • cycle #37 ● REASONING cycle 37 Knowledge → Reasoning → Action

Encoding expert knowledge into language models

Skilled farmers and researchers have accumulated specialized knowledge in their fields through years of experience. We are exploring whether large language models can acquire this expertise and apply it to real-world agricultural tasks. This encompasses a wide range of areas, from autonomous agricultural agents that make decisions in the fields to knowledge graphs that organize scientific literature to support research discoveries.

How LLMs acquire and apply agricultural expertise — flow from farmers and researchers through an LLM into autonomous agents and knowledge graphs, with a task performance radar chart comparing agriculture-tuned vs general LLM

Research Topics

Two current research directions applying LLMs to agricultural knowledge and production systems.

Autonomous irrigation diagram comparing Agent AI and Accumulated Solar Radiation strategies

Autonomous Irrigation

Function Calling Experience Memory Substrate Sensing

LLM agents read sensors, decide irrigation volume, and learn from outcomes over successive growing cycles

Advanced Knowledge Graph System interface showing graph visualization and analysis panels

Knowledge Graph Construction

Literature Mining Entity Extraction Research Discovery

Automated mining of scientific literature to build large-scale knowledge networks for research discovery

You found me!