ADVANCED MATERIALS & PROCESSES | JULY 2026 10 MACHINE LEARNING | AI AI TURNS MICROSCOPY INTO MATERIALS INSIGHTS Researchers at Cornell University, Ithaca, N.Y., developed an autonomous artificial intelligence (AI) platform called EMSeek to streamline materials research. The new tool does its work in mere minutes by identifying key features in a microscopy image, determining the crystal structure, predicting material properties, comparing results with existing scientific literature, and generating a report within a single workflow. “Electron microscopy produces incredibly rich information, but the bottleneck is often turning those images into usable scientific understanding,” explains researcher Fengqi You. “Our goal was to build an autonomous AI platform that helps bridge that gap and makes advanced materials analysis faster, more integrated, and more reproducible.” EMSeek employs an agentic archi- tecture, in which multiple AI agents handle different parts of the workflow and are coordinated by a central system. The platform plans tasks, selects tools, and verifies results, mimicking how a human researcher might approach a complex analysis. The team demonstrated that EMSeek can process a microscopy image into a structured scientific output in just two to five minutes, roughly 50 times faster than conventional expert workflows. The platform was tested across 20 different materials and five tasks typically performed by re- searchers, showing strong performance across a range of conditions. In addi- tion, each step includes checks for consistency and accuracy, helping ensure results are transparent and reproducible. cornell.edu. AI IDENTIFIES MATERIALS SCIENCE RESEARCH TOPICS In a new study, researchers from the Karlsruhe Institute of Technology (KIT), Germany, and their partners have shown how new research ideas can be identified from the growing number of scientific papers now being published in all disciplines. Using artificial intelligence (AI), the team systematically analyzed materials science publications to determine potential new avenues of research. Because materials science is an interdisciplinary field with a strong impact on many technology areas, it has a correspondingly large volume of research papers. However, the findings described in these papers are only useful if relevant trends can be identified. “Our goal is to support researchers in their creative thought processes by shedding light on new avenues of research and opportunities for inter- disciplinary cooperation,” says KIT researcher Pascal Friederich. In their project, the scientists combined large language models (LLMs) with machine learning (ML) methods. The LLMs begin by identifying key terminology and scientific concepts in the journal articles. This information is used to generate a concept graph, a knowledge network in which each keyword forms a node. A second ML model connects nodes when terms are mentioned together particularly often in scientific papers. The model does this by analyzing how links between terms change over many years. When certain concepts are linked with increasing frequency, this can be an indication that a new field of research is emerging. Conversely, a decrease in the number of links can be an indication that certain topics are losing relevance. kit.edu. An AI-generated knowledge network of technical terms reveals trends in materials science research. Courtesy of KIT. Part of a one-click reference-patch framework for universal electron microscopy segmentation. Courtesy of Science Advances, 2026, doi.org/10.1126/sciadv.aed0583.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==