Software Engineer

Benefit Mint – Admin Panel is a comprehensive web-based platform developed to manage and configure complex employee benefit structures within an organization. This admin panel is built using React, Redux, TypeScript, and Java Spring Boot, providing a dynamic and scalable interface for HR administrators. It enables effortless management of benefit types, eligibility criteria, approval workflows, and supportive data fields. The system supports advanced configurations such as dynamic approver logic, conditional supportive fields, and entitlement validation, making it highly adaptable to various organizational policies. With a focus on usability and flexibility, the admin panel streamlines benefit operations and enhances administrative efficiency. Seamless integration with a microservices backend ensures modularity, maintainability, and secure data handling.

Associate Software Engineer

Track - SU is an advanced, cross-platform desktop application designed for online employee monitoring. Built using React, TypeScript, Firebase, and ElectronJS. Track - SU combines innovative technology with user-friendly design to deliver an efficient and secure solution for managing remote and in-office teams. It offers seamless real-time monitoring capabilities, enabling organizations to enhance productivity and streamline workforce management. It runs smoothly on Windows, macOS, and Linux, ensuring accessibility for diverse teams. It Tracks employee activity, including breaks and idle times, and captures real-time screenshots as proof, empowering managers to make well-informed decisions. Track - SU Leverages React for an intuitive user interface, TypeScript for enhanced reliability, Firebase for robust backend services, and ElectronJS for desktop functionality.

Software Engineer - Intern

Fashion Nova is a sleek and modern web template designed for clothing and fashion websites. Developed using Next.js, Bootstrap, and TypeScript, it combines cutting-edge technologies with a clean, responsive design to deliver an exceptional user experience. The template is powered by Storybook, providing a robust component playground that simplifies UI development and testing. It allows developers to build and showcase UI components in isolation, streamlining the design workflow. Additionally, it ensures high performance, server-side rendering (SSR), and seamless navigation for a smooth and efficient user experience.

Data Scientist - Intern

SwingZen is an AI-powered golfer training mobile application designed to help users refine their golfing techniques through advanced video analysis. Built with cutting-edge technologies and enhanced with YOLOv7, the app leverages object detection and pose estimation models to provide precise insights into a user\'s swing mechanics. Implements advanced techniques to detect and mitigate the effects of poor lighting conditions, ensuring accurate performance under various environments. Features efficient video trimming algorithms to focus on relevant segments of a golf swing and streamlines data processing for optimal analysis. Integrates ONNX models for cross-platform compatibility. Utilizes model quantization and pruning to enhance speed and reduce computational load. Explores lightweight solutions like SSD Lite MobileNet v3 for mobile optimization and flexibility. SwingZen combines AI innovation with practical functionality, making it an indispensable tool for golfers seeking to enhance their skills and performance through smart, data-driven training.

Data Scientist - Trainee

In this research, a solution to the cold start problem was proposed by combining the collaborative filtering algorithm with the Jaccard similarity index. The cold start problem arises when a recommendation system struggles to make accurate predictions due to a lack of sufficient data. This typically occurs with New Users and New Items. Collaborative Filtering: Leverages the preferences of similar users to predict recommendations for new users or items and Jaccard Similarity Index: Measures the similarity between users or items by comparing the overlap in their interaction sets, enhancing the system\'s ability to make accurate recommendations even with limited data. This approach improves recommendation accuracy and mitigates the cold start problem by efficiently utilizing sparse datasets and finding meaningful patterns in user-item interactions.

Freelance Machine Learning Engineer