As a Software Development Engineer II at Amazon, I spearheaded the development of an innovative continuous IP violation discovery system leveraging cutting-edge machine learning and semantic search techniques. This groundbreaking solution enabled the proactive identification and removal of counterfeit products from Amazon's vast product catalog on a large scale.Key Achievements:- Architected and engineered the end-to-end system, integrating state-of-the-art models like CLIP for similarity search and optimized embedding generation, delivering substantial cost savings through efficient model deployment.- Implemented a high-performance discovery pipeline ingesting risk signals, performing efficient nearest neighbor searches on a massive vector index, and surfacing potential infringements to detection systems while significantly reducing memory footprint.- Optimized ML models for batch processing, enabling a substantial increase in throughput and driving a significant rise user productivity.- Collaborated cross-functionally to seamlessly integrate the system into existing IP violation detection workflows, substantially reducing manual efforts and driving considerable cost savings.Through this system and other performance optimization initiatives, I significantly enhanced Amazon's ability to identify and remove counterfeit products at scale, safeguarding intellectual property rights and maintaining a trustworthy shopping experience for customers.

As a SDE at Amazon, I played a pivotal role in developing advanced systems for intellectual property protection and violation detection, leveraging natural language processing (NLP) and automated query generation techniques.Key Achievements:- Architected a high-performance rule execution system leveraging advanced Lucene-based indexing with custom NLP analyzers, capable of processing at 10K+ TPS and evaluating complex rules in real-time for IP infringement detection while optimizing rule storage and retrieval.- Developed a sophisticated query generation algorithm to create targeted Lucene queries, reducing the number of rules requiring full evaluation and improving overall system performance.- Engineered a flexible rule syntax supporting nested conditions and fuzzy search operations, ensuring accurate IP infringement detection across diverse product attributes.- Engineered an automated AWS Step Functions workflow to detect illegal use of major sports brand IPs, generating keywords from brands, creating OpenSearch queries to process the entire catalog, and integrating a UI for efficient manual audits, identifying thousands of infringing products while saving significant person-hours.Through these innovative NLP-based techniques, I played a crucial role in enhancing Amazon's IP protection measures, enabling efficient detection and removal of infringing products, and safeguarding intellectual property rights across Amazon's vast product catalog.

As an SDE Intern at Amazon, I developed a full-stack search configuration playground for our search service clients, enabling efficient testing and customization of search parameters.Key Contributions:- Built a responsive and scalable solution with a user-friendly React frontend and RESTful API backend.- Enhanced the onboarding experience for new clients through an intuitive interface for configuring and comparing OpenSearch specific search configurations.Through this project, I gained valuable experience in full-stack development and contributed to improving the client experience for Amazon's search services.

Software development for a simulation based AI course.

Built websites and an iOS application to view 3D models in Augmented and Virtual reality.

Worked on a VR training device for tracking body exercises using a Microsoft Kinect. Built the gesture builder and recognizer software engines for the equipment.