Spring 2025 CS Capstone Projects
Sample projects from CS 490:
1) Carbon in Context
Students developed Carbon in Context, a tool that translates greenhouse gas emissions into relatable comparisons to improve public understanding of climate data. By converting figures like gigatons of CO₂ into everyday equivalents, the tool helps prevent greenwashing and miscommunication. The project was supported by Project Drawdown, a leading climate solutions organization. To see the tool in action, visit: https://www.carbonincontext.com/.
2) NuffCash Model
Students contributed to Nuff Cash, a fintech project sponsored by Qualcore Innovations, by developing a machine learning model to enhance the existing mobile app. Using algorithms like XGBoost, Random Forest, and Isolation Forest, the model analyzes user transaction behavior to detect spending patterns, flag anomalies, and generate alternative credit scores—particularly for unbanked or underbanked individuals. Integrated via API, the model delivers real-time, personalized financial insights to help users budget and manage their finances. The project supports Nuff Cash's mission to improve financial inclusion by digitizing cash transactions while preserving their flexibility.
Sample Projects from CS 690:
1) Project: Cloud Pathology and AI
Sponsor: Connection Loops - Gaurav Kalele
Team: Haebin Lee, Rucha Maslekar, Ananya Agarwal
This project's goals are to transform pathology laboratories by integrating advanced AI-driven technologies, managing the end-to-end patient journey within a secure and compliant cloud-based ecosystem, and serving over 600 laboratories and processing more than 2 million patient records annually. The primary motivations include alleviating medical workloads by automating repetitive tasks, enhancing lab efficiency through improved accuracy and streamlined operations, and providing diagnostic solutions by offering personalized lifestyle recommendations and actionable insights for patients. By achieving these goals, Cloud Pathology AI aims to optimize the workflow in pathology labs and significantly improve patient care and outcomes.
2) Project: Doctor Right
Sponsor: Doctor Right - Nathan Stott
Team: Suresh Selvadurai, Chaoping Zhong, Tingxuan Lin
The "Doctor Right" system uses machine learning to recommend cost-effective, high-quality doctors tailored to patient needs. It employs two models: the Patient Model clusters patients based on features like demographics and diagnoses, while the Doctor Model predicts claim amounts using doctor experience, specialties, and treatment outcomes. Doctors are ranked based on predicted costs and cluster-specific experience to ensure effective recommendations. Utilizing an XGBoost classifier, the system achieves an F1 score of 0.66, reflecting a performance capable of driving meaningful changes and informed recommendations while leaving room for optimization. Future improvements could include incorporating more diverse patient and doctor attributes and refining predictions for even more personalized healthcare recommendations.
3) Project: Snaplogic LLM Agent
Sponsor: Snaplogic - Greg Benson
Team: Heran Zhan, Qianru Wei, Yin Hu
SnapLogic’s integration platform revolutionizes data and application connectivity. Our team has integrated an LLM Agent Framework with snapLogic pipelines, extending the automation and sophistication of the GenAI Snaps developed previously. This poster highlights the SnapLogic LLM Agent’s structure, key use cases, and the Agent Log Visualizer Snap for tracking task execution.
The framework is versatile, supporting multiple LLMs like Anthropic Claude, OpenAI GPT, and Google Gemini, and diverse tools. Advanced features like the memory unit, ReAct (Reasoning and Acting) design pattern, and multi-agent communications ensure superior performance and adaptability.