Projects

Chronicle anything, beautifully. A minimal, storytelling-first tool to create concise year-by-year timelines for personal journeys, projects, or anything that unfolds over time. With export and oEmbed-powered embeds.

An AI-powered Personal Operating System (pOS) that turns goals into habits and habits into actionable tasks through structured accountability and behavioral science.

A curated, AI-assisted social network for intentional connections — helping people find co-founders, collaborators, and meaningful professional relationships based on goals and values.

Developed a recommendation engine using ML techniques (TF-IDF, cosine similarity, semantic search) for personalized movie, TV show, and book suggestions and deployed the app on iOS and Android stores. Built an over 300,000-item SQL database with tools like SQLModel and SQLAlchemy for high concurrency and efficiency.

Built a private, family-first social platform designed to help families and close friends stay connected through shared moments, events, and memories — without feeds or algorithms.

Developed a production-ready video assistant web app, EVA, as a part of capstone project which offers hybrid approach to video editing. Utilized ffmpeg for robust video processing. Designed with a user-centric approach, EVA can be used by content creators of all levels. It offers a streamlined experience in video editing and audio processing, requiring no prior editing expertise. Integrated open-source machine learning models into workflow to enhance features: improve audio, give multi-layered feedback and condenses video to key moments.

Built an open-source system to enable automation for applications by translating natural language queries to programmatic actions. Designed vector-store and parameter-extractor components for efficient similarity search using LLM and NER for query processing. With over 1500+ installs, it supports use cases like API interfaces, chatbots, automated task systems, and voice-controlled apps.

WildLens is a machine learning model for fauna recognition, for a non-invasive approach to continuous wildlife census. This system can be invaluable for automating species identification, aiding wildlife researchers, and contributing to biodiversity conservation efforts. Our model trained on large datasets of around 60,000 images while the ethical considerations were prioritized.

A real-time geospatial experiment that visualizes the gap between headline-driven negativity and long-term global progress — built to counter the social media filter effect.

Developed a video editing application that enables text-to-video edits by analyzing and executing text-described actions. Implemented a backend in Python and FastAPI to process text inputs, detect actions, extract parameters, and determine edits.

Developed an automated software solution, simplifying the tasks of merging and reformatting data from diverse sources. This resulted in the creation of an efficient and current departmental database. Streamlined the navigation process from the old database format to a new, efficient format that provides enhanced and relevant information for each member of the department. Utilized PyZotero and web-scraping techniques to integrate data and references, enhancing the accuracy and credibility of the department's database content. Contributed to enhancing departmental efficiency by providing a data-driven solution that enabled faster access to accurate information, ultimately supporting better decision-making processes.

PragYantra is a simple software project to simulate a humanoid robot with vision, hearing, speech, and memory functionalities. I wanted this project to create a flexible platform for experimenting with artificial intelligence and human-machine interaction. It can work fully offline (may require a stronger device for faster inference) and online for low-end devices.

A Face Recognition and Privacy Analysis Tool, a software application developed with the primary goal of providing face and object recognition capabilities. The inspiration for this tool is the growing concern regarding data privacy and the potential misuse of facial recognition technology. Goal is to build lot of features on top of this face recognition tool.

Recipient of the Best Emerging Technology Award at SpartaHack. Developed a fully operational, full-stack educational web application utilizing the Django framework. Empowers personalized learning with a multitude of components such as custom quizzes and flashcards tailored to students educational requirements. This provides a range of complementary services that enhance personalized learning products, equipping students with comprehensive tools essential for success throughout their academic journey. Currently in the development phase, our goal is to construct highly customized AI models designed to assist students across different grades, subjects, and formats.

VidCaptio is a free video captioning software. Multi-Language Captioning allows users to add captions in multiple languages to their videos. Users have the option to choose which languages they want to add captions for. Uses OpenAI's Whisper to accurately transcribe speech from videos for captioning. Utilizing ffmpeg, VidCaptio processes videos for captioning, ensuring high-quality output.

PyOrcid is a Python package and API client designed to simplify interactions with the ORCID API. ORCID (Open Researcher and Contributor ID) is a nonprofit organization that provides unique identifiers to researchers, ensuring their work is accurately attributed and discoverable. PyOrcid enables developers to seamlessly integrate ORCID functionality into their software, allowing users to collect, track, and sync their publication materials, research activities, and other related information.

Developed an innovative variation of the classic Wordle game, featuring a dynamic and user-engaging interface including a daily leaderboard. Utilized Python and Flask for backend development, handling game logic and server-client interactions efficiently. Integrated Ajax for smooth and asynchronous data exchange, keeping the gameplay fluid and responsive. Implemented user authentication features, requiring players to log in, thus personalizing the gaming experience and enabling score tracking.

Developed a musical machine that utilized card input (with rows mapped to musical notes) to enable the various components to play. Designed a comprehensive UML class diagram, incorporating associations, inheritance, and attributes, as a preliminary step to development. Implemented advanced programming concepts such as Tweening, Visitor and Observer patterns, Polymorphism, Adaptation, Inheritance, and Composition. Successfully integrated and adapted the animated software into another program, to create an animated movie.

Created a multi-level game called Bug Squash involving various types of bugs attempting to infect programs. Implemented agile software development practices along with Unit-Testing practices, dynamic Animations, OOP, and concepts of Visitor and Observer patterns, Polymorphism. Contributed to a winning team in a Level 3 design contest organized by the CSE department.

Developed a fun program involving a heavy animated aquarium that can be populated. This project involves concepts like file saving/loading of formats, Refactoring, Inheritance, and Unit testing.

I created an interactive piano entirely using HTML, CSS, and JavaScript. This project not only highlights my web development skills but also demonstrates my ability to build beautiful and mobile-friendly front-end applications.