Kolkata metropolitan area, West Bengal, India
I'm a Software Engineer and B.Tech in Information Technology student at the Indian Institute of Information Technology (IIIT) Allahabad, maintaining a CGPA of 8.57. I am passionate about building intelligent, efficient, and scalable software solutions that solve real-world problems. During my project internship at Oracle, I developed a machine learning-based de-identification pipeline to anonymize Personally Identifiable Information (PII) in medical images. By architecting a hybrid detection system using GPT-40 and YOLOv5, I improved the reliability of tattoo detection on over 700 clinical images. My contributions to the Note-Eval Factory tool also led to a 3x reduction in token usage and a 25% boost in evaluation consistency. My experience also includes full-stack development. I independently developed projects like EcoBloom, an interactive campaign management platform recognized in the Global Top 100 of the Google Solution Challenge '24, and CodeX, a scalable coding contest platform. These projects demonstrate my ability to build complete, user-focused applications from the ground up. My work spans both machine learning and full-stack development, with a strong focus on robust architecture and creating seamless user experiences. I am proficient with a range of technologies including C++, Python, JavaScript, ReactJS, NodeJS, Express.JS, MongoDB, Firebase, and Docker. I am driven by a deep curiosity for technology and thrive in fast-paced environments where I can tackle complex challenges and contribute to impactful products. I'm always excited to collaborate on projects involving machine learning, full-stack web development, and data-intensive applications.
Built production-grade backend services for One Medical, enhancing authentication, account security, and session management across web and mobile applications. Delivered end-to-end features across multiple codebases, collaborating with security, frontend, mobile, and infrastructure teams to improve platform reliability, strengthen user protection, and enable seamless production rollouts with zero customer impact.
Developed the core ML pipeline for an automated de-identification system to anonymize PII in medical images before EHR integration. Built a hybrid detection pipeline using Python, GPT-4o, and YOLOv5, improving detection across 700+ clinical images and reducing misclassifications by 40%. Enhanced the Note-Eval Factory by integrating GPT-based evaluation pipelines, achieving a 3× reduction in token usage and a 25% improvement in evaluation consistency.
As Coordinator, I led the UI design for the OpenCode website and mentored 50+ freshmen across 4 graphic design workshops.