ClimateAI - Multi-modal System for Climate Analysis
Sheikh, Muhammad Umer
Sheikh, Muhammad Umer
Author
Supervisor
Department
Machine Learning
Embargo End Date
2026-05-30
Type
Thesis
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
The increasing complexity of climaterelated challenges necessitates the development of intelligent, multi-modal AI systems capable of reasoning across diverse data sources. This research presents a Multi-Modal-Climate AI framework that develops an agentic flow to enhance decision-making in climate science. The framework features a finetuned language model trained on domain-specific knowledge, enabling dynamic intent recognition and adaptive tool selection for climate related analysis. The central hypothesis of this study is that an AI-driven, multi-modal system with an intent-based tool orchestration mechanism can improve the accuracy and efficiency of climate-related decision-making. To evaluate this hypothesis, the framework employs a structured methodology where user queries are analyzed, classified, and routed to the most appropriate analytical tool. These tools encompass retrieval-augmented generation, expert recommendations, and specialized classifiers tailored for environmental applications. The model is optimized through knowledge transfer from largescale models and finetuned with domain-specific datasets to enhance its contextual understanding of climate-related issues. The key contributions of this research include the development of a novel AI-powered orchestration framework for climate applications, an enriched dataset tailored for climate and regional inquiries, and an adaptive system capable of integrating multiple analytical tools. Experimental results demonstrate that the proposed framework significantly enhances climate-related analyses by improving response accuracy and contextual relevance. This work lays the foundation for advanced AI-driven climate intelligence systems, contributing to the fields of climate monitoring, disaster assessment, and ecological research.
Citation
Muhammad Umer Sheikh, “ClimateAI - Multi-modal System for Climate Analysis,” Master of Science thesis, Machine Learning, MBZUAI, 2025.
Source
Conference
Keywords
Large Language Model (LLM), Agents, Climate-and-sustainability, Synthetic-dataset, Mulitmodal-agent