Research
Accelerating Materials Discovery with AI-Powered Electron Microscopy Analysis
Accelerating Materials Discovery with AI-Powered Electron Microscopy Analysis
Prof. Fengqi You, Cornell University
Using AI to transform atomic-level images to scientific insights
The Challenge
Electron microscopy (EM) allows scientists to observe materials at the atomic scale, providing critical insights for technologies ranging from advanced batteries and catalysts to semiconductors and next-generation manufacturing. However, transforming electron microscopy images into useful scientific knowledge remains a slow and complex process.
Researchers often must complete multiple specialized steps, including identifying structures within images, reconstructing crystal arrangements, predicting material properties, and reviewing relevant scientific literature. These tasks typically require significant expertise and can take days or even weeks to complete.
While artificial intelligence has begun to assist with individual parts of the process, most existing tools are designed for a single task and do not work well together. As a result, scientists still spend substantial time manually coordinating complex workflows.
The Approach
Researchers led by Cornell University’s Fengqi You developed EMSeek, an AI-powered platform designed to automate the entire workflow from electron microscopy image to scientific insight.
EMSeek combines multiple specialized AI systems into a single coordinated platform. The system can identify and segment structures within microscopy images, reconstruct crystal structures, predict material properties, retrieve relevant scientific literature, and verify that results are physically consistent.
Large language models serve as the platform’s coordinator, automatically determining which tools are needed, managing the workflow, and integrating results into a single analysis. The system also tracks the origin of every result and generates audit-ready reports, providing transparency and confidence in its findings.
By combining these capabilities into a unified framework, EMSeek helps researchers move from raw microscopy data to actionable scientific understanding with minimal human intervention.
Why Empire AI Matters
Developing and testing a system like EMSeek requires significant computing power. Researchers must build, deploy, and evaluate an integrated AI-agent workflow, rather than only individual models. This includes specialized vision and prediction models, LLM-based coordination, scientific tool integration, and validation across diverse microscopy images, materials systems, and tasks.
Empire AI provides the advanced computing infrastructure needed to support these large-scale experiments. Access to high-performance computing enables researchers to develop more sophisticated models, evaluate them across diverse datasets, and refine the platform’s ability to automate complex scientific workflows.
These resources allow academic researchers to pursue ambitious AI-driven scientific discovery projects that would otherwise be difficult to achieve at scale.
What Researchers Are Learning
Results demonstrate that AI can do more than assist individual scientific tasks—it can coordinate entire research workflows.
Across multiple materials systems and analysis challenges, EMSeek delivered faster and more accurate results than many existing approaches. A complete analysis can be performed in as little as two to five minutes per image, approximately 50 times faster than traditional expert-driven workflows.
The research also highlights the growing role of AI agents in science. Rather than functioning as standalone tools, AI systems can increasingly act as intelligent collaborators that connect data analysis, prediction, validation, and knowledge discovery into a unified process.
Potential Impact
By dramatically reducing the time required to analyze electron microscopy data, EMSeek could accelerate discoveries across materials science.
The platform may help researchers more quickly identify promising materials for clean energy technologies, advanced manufacturing, electronics, and other critical applications. Faster analysis can shorten the path from scientific observation to practical innovation, helping researchers focus more time on discovery and experimentation.
More broadly, the project demonstrates how AI can transform scientific research by automating complex workflows while maintaining transparency, rigor, and scientific accountability.
Impact for New York
Advances in AI-driven materials science can strengthen New York’s leadership in advanced manufacturing, semiconductors, energy technology, and scientific research. Projects such as EMSeek help equip researchers with cutting-edge tools for discovery while training the next generation of scientists working at the intersection of artificial intelligence and materials engineering.
By supporting innovations that accelerate research and development, Empire AI helps position New York as a leader in both AI-enabled science and the industries that depend upon it.