Developed a DeepLoc2.0-based ML model with advanced neural networks and uncertainty estimation, improving prediction accuracy. Built a full-stack web app integrating an ML model for explainable protein localization predictions with uncertainty measures. Addressed significant accuracy drops on independent datasets by integrating uncertainty quantification methods like deep ensembles and Monte Carlo dropout, ensuring robust and reliable predictions.

Led a team in creating an Interactive Book Reader with Augmented Reality (AR), revolutionizing digital reading. Developed a unified e-reading platform accessible across web and app, enhancing user engagement with features like cross-platform syncing and APIs for text-to-speech, dictionary, and real-time translations. Integrated AR seamlessly using Unity, enhancing the digital reading journey with real-world AR models. Committed to ongoing platform refinement to redefine the way literature is consumed.

Innovated an accessible Whack-a-Mole game tailored for visually impaired individuals, bridging a significant accessibility gap in the gaming industry. Utilizing Unity for game development, the focus was on implementing object-oriented programming (OOP) concepts in C# for an efficient and robust design. Multiple APIs were integrated to ensure the game's accessibility for visually impaired players, enhancing their gaming experience. This project advanced to the final stages of the 'Idealize' competition, recognizing its innovative approach and impact on inclusive gaming.