Faculty and graduate students in the University of Maryland Institute for Advanced Computer Studies (UMIACS) are heavily involved in a new seed funding program that incentivizes research and innovation focused on artificial intelligence. The grants are administered by the Artificial Intelligence Interdisciplinary Institute at Maryland (AIM), a collaborative hub that the university launched last spring.
Of the 22 research projects receiving a share of the $1.5 million in AIM seed funding announced on May 28, UMIACS researchers are active in 10 of them. Projects involving UMIACS faculty and students are:
Cross-College Collaborative Awards ($100,000-$300,000)
CommunityTwin: A Digital Twin for Enhanced Decision-Making in Public Health
Vanessa Frías Martínez, Louiqa Raschid and Philip Resnik are working with others to develop a highly detailed, yet adaptable, set of models of a community known as a “digital twin” to assist policymakers in designing personalized, localized interventions for pressing community challenges. The system will integrate deep learning, human mobility data and fine-tuned large language models to simulate real-world behaviors and community mental models.
AI-Driven Sensor Fusion for Arctic Sea Ice Mapping
Christopher Metzler is working with others to develop advanced AI-driven photogrammetry and sensor fusion techniques to enhance long-term coastal monitoring efforts. The team will identify scientifically significant Arctic Sea ice regions and then develop low-cost and accessible sensing platforms to survey the areas and form high-resolution 3D reconstructions of the ice above and below the sea surface.
Human-Centered AI for Translating Classical Chinese
Marine Carpuat is part of a team that will design human-centered AI technology to reliably translate classical Chinese and integrate AI literacy into the humanities curriculum.
AI for Motor Learning in Instrumental Education
Cornelia Fermüller and Irina Muresanu are developing AI-driven tools to help students master bow techniques for string instruments. They will develop algorithms that analyze video and audio inputs to provide precise, actionable feedback to students.
Theory and Metrics for Law-Informed Attribution of Human-Generative AI Liability
Furong Huang and graduate student Tin Nguyen are working with others to develop a framework to determine how much liability companies that deploy AI systems should bear compared with users who may be harmed by the technology. The framework will be informed by the “joint and several liability” principle from tort law and the “path model of blame” from psychology.
Blind Motion Prediction for Robotic Planning
Hernisa Kacorri and Abhinav Shrivastava plan to collect and share a novel multimodal motion benchmark dataset tailored to the movements of individuals who are blind. Using this dataset, they will evaluate the ability of generative AI models to predict how blind individuals interact with objects, and assess the potential to develop robots that can interact more safely and smoothly with visually impaired people.
Co-Learning Code and Mind: Integrating AI With Transformative Social and Emotional Learning for Diverse Youth Through Tangible Robotics
Huaishu Peng is working with others to engage middle school students alongside their neurodiverse peers in the hands-on co-designing, co-building and co-programming of robots. They will partner with two nonprofit organizations in Montgomery County for this project: the AOE Robotics Club and the Special Education Equal Development Society.
Individual Postdoc and Student Awards (up to $5,000)
Introducing Spatial Speech Context in Wearable Large Language Models
Graduate student Ayushi Mishra, who is advised by Nirupam Roy, plans to develop a novel system architecture that incorporates spatial speech understanding into large-language models, enabling contextually aware and adaptive applications for wearable technologies. Spatial speech integration elevates wearables into intelligent companions—enhancing productivity, safety and convenience in many scenarios.
Multimodal Volumetric Imaging and Material Identification of Multilayered Objects
Graduate student Harshvardhan Takawale, who is advised by Nirupam Roy, plans to develop a multimodal volumetric imaging framework that integrates millimeter-wave radar and ultrasonic transducers for high-fidelity 3D reconstruction and material identification of multilayered heterogeneous objects. Their work will address critical focus areas in AI-driven sensing, non-destructive testing and medical imaging, where accurate subsurface inspection or complex material characterization is vital.
AIM Fellow Program
Establishing an AI Technology Policy Hub of International Standing at UMD
Katie Shilton and others plan to bring Lee Tiedrich, an internationally renowned scholar of AI, to UMD for the one-year AIM Fellow residency. Tiedrich will engage with the College of Information’s Tech Policy Hub, which encourages policy practitioners, industry leaders and AI governance scholars from the D.C. metro area to connect and share knowledge on the design, implementation and enforcement of AI policies at all levels.