èßäapp

By Dave DeFusco 

At the Spring 2025 Graduate Computer Science and Engineering Research forum, the future of technology wasn’t merely on display—it was actively being built. In their research presentations, Katz School students showcased the kind of ingenuity, technical depth and ethical awareness that are not only rare in graduate education but transformative in the real world.

With many projects accepted at peer-reviewed conferences and others poised for publication, this year’s presentations offered a glimpse into a community of emerging engineers and data scientists who are fluent in both theory and practical application. These students are more than programmers or problem solvers—they are builders of tools that matter. 

“Their work reflects the Katz School’s distinctively interdisciplinary and human-centered vision for computer science and engineering, grounded in both technical mastery and a deep awareness of the social, psychological and ethical dimensions of technology,” said Dr. David Li, the symposium’s host and director of the M.S. in Data Analytics and Visualization.

A highlight of the event was the Best Research Project Award, given to Nikkat Afrin, a student in the M.S. in Data Analytics and Visualization. Her project, “SigmaCam: Exact Decision Boundary Extraction for DNNs with Smooth Nonlinearities,” tackled a major challenge in the field of artificial intelligence—understanding how deep learning models make decisions.

In simple terms, Afrin developed a new way to clearly visualize where and how an AI model draws the line between different categories. This is especially useful when models use smooth activation functions—like Sigmoid or SiLU—which are common in real-world applications but harder to analyze. Her algorithm makes these decision boundaries visible and accurate, helping researchers and users trust AI systems more. The work pushes the boundaries of model transparency and could be vital in areas like healthcare or finance, where understanding model behavior is essential.

Another Best Research Project Award went to Hyeonwook Kim, a student in the M.S. in Artificial Intelligence, for his project, “RetinFormer: Augmented Vision Transformer for Diabetic and Hypertensive Retinopathy Classification.” Retinopathy—eye damage caused by diabetes or high blood pressure—can lead to blindness if not caught early, but diagnosis is time-consuming and requires expert knowledge. Kim’s solution is an AI model that can automatically analyze retinal images and classify them with over 92% accuracy, reducing the time needed for diagnosis and making screening more accessible.

Using a sophisticated Vision Transformer architecture, RetinFormer can distinguish between diabetic retinopathy, hypertensive retinopathy and normal eye images. This is not just an academic exercise—it’s a potential tool for improving public health, especially in communities where access to specialists is limited.

Not all research has to be about complex algorithms—some is about improving everyday life. Xiaohang Liu, an artificial intelligence master’s student, won a Best Research Presentation Award for his clever mobile app, PackMind+. We’ve all forgotten something while packing for a trip. PackMind+ makes sure that never happens again. Designed for iOS and Android, the app uses intelligent suggestions to recommend what to pack based on your plans—whether you’re heading on a snowboarding trip or just doing a grocery run.

Users can customize their lists, collaborate with others and receive reminders so they never miss an item again. What makes it stand out is its thoughtful design and user experience—simple icons, drag-and-drop lists and subtle animations make packing not just easy but enjoyable. It’s a great example of how technical skills and human-centered design can work together to solve everyday problems in delightful ways.

The other Best Research Presentation Award went to a team of Data Analytics and Visualization master’s students—Anupam Semwal, Arunesh Kumar Rai, Priya Tinna, and Aswin Karthik Panneer Selvam—for their project on predicting cryptocurrency prices using Time Series Foundation Models (TSFMs).

Cryptocurrency is known for its wild price swings, making accurate forecasting incredibly difficult. The team analyzed how new AI-based time series models like Chronos, TimesFM and Tiny Time Mixers compared with more traditional tools like ARIMA and GRU. They applied their analysis to four major cryptocurrencies—Bitcoin, Ethereum, Solana and Binance Coin—using over two years of trading data.

Their work goes beyond finance. It shows how AI tools originally built for language tasks can be adapted to work on time-series data—like stock prices or weather patterns. The team created a detailed forecasting pipeline and offered insights into how model choice, data normalization and prediction window size can dramatically affect accuracy.

Other presentations included:

• Smart Emergency Detection Watch Using Real-Time Sound Classification — Dhwani Patel (AI), Jestin George (AI), Kousik Gunasekaran (Cybersecurity)
• Mind the Gap: Integrating Psychological Principles for Effective and Ethical AI Development â€” Emmanuel Olimi Kasigazi (DAV)
• Adaptive Deep Learning with Batch Feature Re-Engineering and Differential Dynamical Systems — Ruixin Chen (AI)
• Connectlift Mate: An AI-Powered Chrome Extension for Smarter Job Networking on LinkedIn — Ashish Rogannagari (DAV)
• Cross-Lingual Text Augmentation: A Contrastive Learning Approach for Low-Resource Languages — Hang Yu (AI), Ruiming Tian (AI)
• AI-Powered Fraud Detection in Financial Transactions — Nikki Rastogi (DAV), Akshit Arora (DAV), Ashish Rogannagari (DAV), Sreevani Siddareddygari (DAV)
• SST-EM: Advanced Metrics for Evaluating Semantic, Spatial and Temporal Aspects in Video Editing â€” Varun Biyyala (AI)
• A Dynamic Framework for Optimizing Reward Policies in the Sharing Economy â€” Cheng Li (DAV)
• The Global Decline in Birthrates: Implications for Demographic Shifts, Economic Stability and Public Health — Dhwani Patel (AI), Saugat Sijapati (DAV), Nikkat Afrin (DAV)
• CarbonCultivate AI: A Proof-of-Concept for AI-Driven Soil Carbon Optimization — Yash Negi (AI)
• Empowering Institutional Intelligence Through Scalable Azure-Databricks Architecture and Integrated Dashboard Analytics — JK Vishwanath (DAV), Bhavitha Bojja (DAV), Nikkat Afrin (DAV), Sreyash Mudiam (DAV)
• Forest Surveillance Using Acoustics and Convolutional Neural Networks using Low-Powered SoC Device — Dhwani Patel (AI)
• SmartCanvas: AI-powered Graduate & Undergraduate Chatbots for the Shevet Glaubach Center for Career Strategy and Professional Development — Mapalo Lukashi (DAV), Dheeraj Shankar (DAV), JK Vishwanath (DAV), Chaitanya Devarshi (DAV)
• Building a Global Rare Disease Information Hub to Advance Knowledge, Diagnosis and Treatment — Naveen Khetarpal (BME), Prashant Soni (AI)
• R.E.D. Model: Insider Threat Detection AI System — Geetha Avagadda (Cybersecurity)

“Together, these projects form a powerful mosaic of innovation,” said Dr. Ming Ma, an assistant professor in the Graduate Department of Computer Science and Engineering. “Katz School students are using intelligence—both artificial and human—to solve some of the world’s most intractable problems. Each project is a step forward but together, they are a leap.”