OpenBWC: Analyzing Police Body-Worn Camera Footage uses advanced Machine Learning and Natural Language Processing to enhance police transparency and accountability.
Create a framework for detecting and analyzing patterns in body-worn camera footage. Key challenges included extracting meaningful interactions while prioritizing ethical data handling and minimizing bias, all aligned with improving training effectiveness and fostering community trust.
WhenDoWeEven.ai is a cross-platform calendar web app designed to simplify meeting scheduling. This Progressive Web App (PWA) integrates with Google Calendar, Outlook, and iCal to intelligently recommend optimal meeting times. The app features a custom scheduling algorithm that efficiently parses availability and synchronizes time zones, ensuring accurate coordination for users worldwide. The frontend, built with JavaScript, offers an intuitive interface that enhances the user experience while supporting offline functionality for added convenience, and the backend, built with Python and MongoDB, efficiently handles data processing, calendar parsing, and recommendation logic to ensure accurate and timely meeting suggestions.
To develop an intelligent meeting scheduler that streamlines the coordination of meeting times across different calendar platforms, improving efficiency for global teams through seamless scheduling automation.
Led the creation of an innovative platform for image generation with user interaction and creative output. This project focused on integrating model training and image uploads to enhance user experience.
Develop a platform that simplifies image generation. My goal was to improve accessibility to AI tools while ensuring functionality and user engagement. This initiative not only met its goals but also set the stage for future advancements in user-centric AI applications.
Developed as part of the Computer Science M.S. Capstone requirement at RIT together with the Rochester (NY) Police Department.
Help law enforcement better understand police reports using symbolic AI and natural language understanding. Police reports often contain important details in narrative form, but those details can be difficult to analyze because they are written as unstructured text. This system takes those narratives, extracts important information such as people, locations, objects, actions, and incident types, and organizes them into a structured ontology. This ontology shows how different concepts are connected, such as a weapon being related to violence or an incident being connected to a location. It can then use reasoning rules to infer deeper context, such as whether an incident may be high-risk. A major part of the project is also privacy protection, so sensitive information is redacted before analysis.