REU 2025

Hello! My name is Hannah and I am a rising senior at the University of Massachusetts, Amherst. I am majoring in computer science. This is my first research experience and I am very excited about the opportunity. This summer, I am working with Kunyang Li and Dr. Mubarak Shah on DistillScope, analyzing the scope and limits of diffusion-based distillation methods. Distillation is the process of creating synthetic data with the goal of scaling down large datasets. In this project, we aim to expand existing evaluation methods on optimization-based distillation, like distribution or gradient matching, to diffusion-based distillation methods. We aim to uncover a better understanding of this method, for instance how robust the distilled data is when mixed with real data, and what the learned distributions are sensitive towards. After performing these evaluation methods on diffusion-based distillation, implementing additional metrics of evaluation that are specific to the diffusion method will follow. Feel free to reach out with questions at hcarey890@gmail.com.

Hi! My name is David Gardner, and I am a rising senior at Boston University studying Electrical Engineering and Math. This is my first exposure to computer vision, and I am most interested in its applications to climatetech. This summer, I am working on “Slow-Fast Weights for Long-Term Video Understanding” under Rajat Modi and Dr. Rawat. This project explores applying short-term and long-term memory weights, focusing on different timescales of a 3-hour video for video understanding. Instead of using conventional deep learning techniques like back propagation and convolutional neural networks, we use a forward-forward algorithm and slow-fast weights to understand long videos. Please feel free to contact me at davidgardner335@gmail.com.

My name is Matthew Lewis, and I am a rising senior at the University of Central Florida studying Computer Science. I have some experience in machine learning from taking Artificial Intelligence and Algorithms for Machine Learning but I have no prior experience in research. This summer I am working with Dr. Laura Brattain and Melody Halbert on Applications of Foundation Models and Efficient Diffusion Models to Dental Radiography to Assist Dental Treatment Planning. In this project, we are testing different foundation models like MedSAM and MobileSAM on teeth scans to see how well different models can segment teeth. Improving these models and implementing diffusion models to recreate images of the missing teeth are among the problems we'll be tackling. If you have any questions, please feel free to email me at matthew.lewis@ucf.edu

Hello! My name is Eric Li, and I am a rising senior at UC Berkeley majoring in Data Science, having recently switched from Applied Mathematics. With a strong background in mathematics and some programming experience, this REU marks my first formal exposure to machine learning and computer vision—areas I’m excited to explore more deeply. This summer, I’m working under the guidance of Dr. Tian on a project titled Multi-modal Large Language Models for Video Anomaly Detection. Our goal is to identify the limitations of current frame-to-text prompt video data and enhance scene representation by integrating multi-modal embeddings into MLLMs or LLMs. For updates about my journey here at UCF, please check out my weekly presentations and more below! If you have any questions, feel free to reach out to me at ericli88@berkeley.edu.

My name is Felipe Monsalvo, and I am a rising senior majoring in Computer Science at Kean University. This summer, I am participating in the UCF CRCV REU, where I’m working on Project 29: Understanding Diagrams using Computer Vision, under the lead of Dr. Harry Newton, at Leidos. Our research focuses on developing an agentic AI system capable of interpreting complex maritime logistics diagrams, particularly those composed of cartographic symbols, annotations, and layouts by combining computer vision models such as Grounding-DINO, YOLOv8, and SAM1/SAM2. The goal is to build a multi-agent pipeline with a manager AI that validates outputs for integration into GIS platforms like ArcGIS and integrates into navigation and logistics DB. My contributions will involve model training, symbol detection, and geospatial alignment using satellite and map-based inputs. Feel free to reach out if you’d like to learn more about this work at felipemonsalvocs@gmail.com

My name is Syn Nguyen, and I am a rising Senior at California State University, Long Beach majoring in Computer Science. It is an honor for me to get accepted in the Computer Vision REU at UCF, a very valuable experience. During the Summer, I will be working with Dr. Yu Tian on developing a state of art segmentation model, utilizing SAM2, for automatic cancer detection. Feel free to reach out to me at nsyn1312@gmail.com

Hello all! My name is Munish Persaud. I am a rising senior at UCF dual majoring in Computer Science and Biomedical Sciences with a minor in Nonprofit Management. I have deep interests in computer vision and computational biology. This REU represents my first exposure to computer vision and complements my background knowledge of machine learning and large language models exceptionally well. I am currently working independently with Dr. Bedi on "Jailbreaking a Bio-Foundational Model." This project utilizes Evo2, a generative Bio-Foundational model released by Arc Institute that can produce functional and safe protein structures from DNA input. The goal of the project is to present the perspective that, like large language and diffusion models, heightened safety considerations are needed in the domain of Bio-Foundational models. Please see below for updates about the project and progress made. For any questions, advice to share, or opportunities for collaboration, please contact me at munishpersaud.mp@gmail.com.

Hello! I'm David Rivard, a senior at Lake Superior State University majoring in Computer Science with a focus on data science and machine learning. For the past several years I've worked with many computer vision systems and had opportunities to work on projects involving satellite tracking, image processing, and automated detection pipelines. Through my many projects, I've built a strong foundation in computer vision and machine learning. This summer I'll be learning a lot about multi-modal learning by researching under Rohit Gupta, Parth Kulkarni, and Dr. Mubarak Shah. The research project I will be contributing to is Geo-HiPO: Geolocalization via Hierarchical Preference Optimization of Multi-Modal LLMs. The project is about training vision-language models to predict geographic place names using preference-based learning grounded in real-world distances. I look forward to contributing to cutting-edge research in geolocalization. Please contact me at davidrivard4@gmail.com.

Hello! I am a rising junior majoring in Mathematics, with minors in Computer Science and Data Science. I have taken two courses in linear algebra and a first course in real analysis; on the applied side, I have taken multiple courses in data science, including Big Data Analytics Methods and Statistical Foundations of DS and AI II. My academic interests primarily revolve around bringing together my mathematical understanding and practical data science experience to create tomorrow’s state of the art algorithms. With my advisor, Dr. HanQin Cai, my work for this summer will involve working on matrix recovery algorithms in relation to signal processing problems. Building from Dr. Cai’s prior work, my hope is to end the summer with an algorithm that is more memory efficient and capable of recovering signals in more ill-conditioned circumstances than its predecessor.

My name is Neel Shah, and I’m a rising senior at Rutgers University, majoring in Electrical Engineering with minors in Computer Science and Mathematics. This is my second REU experience—last summer, I worked on Machine Learning and NLP at the University of Missouri, which was my first hands-on exposure to ML. This summer, I’m working with Mr. James Benjamin on a project focused on generating dance videos conditioned on audio and images using VQ-VAE and Diffusion Transformers. The goal is to improve the consistency and realism of human motion in AI-generated content. I’m excited to deepen my understanding of generative models and their applications in multi-modal video synthesis.

My name is Ejiro Adams Ubini, and I’m a final-year Computer Science major at the University of Maryland, Baltimore County. I’m deeply passionate about technology, with a strong desire to become a software engineer who leverages artificial intelligence to build smarter, more adaptive applications. This is my fourth summer research internship and my third REU experience. Through these opportunities, I’ve gained hands-on experience in areas like machine learning, computer vision, and real-time systems. This summer, I’m working on the project “Real-Time Detection of School Bullying Using CCTV” under the mentorship of Dr. Yogesh Rawat. The project focuses on using deep learning and video activity recognition models to detect bullying behaviors in real time from school surveillance footage. We plan to curate video datasets by identifying and annotating scenes that capture behaviors like physical aggression, verbal altercations, and normal interactions. Using these labeled examples, we’ll be fine-tuning foundation models for activity recognition – evaluating their performance on full video sequences versus shorter clips. The goal is to build an efficient system capable of detecting bullying incidents in real time, with our findings documented in a structured research paper. For any questions or to get in touch, feel free to contact me at ubiniadams@gmail.com.