Design and development of AI powered recommender systems at scale. Working on enhancing Amazon Personalize service experience for customers with below contributions.Generative AI-Powered Capability for augmented recommendations:- Led the design, development, and launch of a generative AI-powered feature for Amazon Personalize which enabled generating thematic connections between recommended items for use cases such as marketing email headings and carousel titles.- Collaborated with the product team to gather requirements and define scope, owned system design and led a team of developers to deliver the project under tight deadlines.- Worked with the science team to fine-tune prompt engineering and benchmark large language model (LLM) outputs for quality assurance.Personalize Search Feature:- Spearheaded the cornerstone launch of the Personalize search feature, integrating Amazon OpenSearch with Amazon Personalize.- Worked ground up for the project launch from requirement gathering and scoping to design, implementation, testing. Facilitated effective communication across AWS Personalize and OpenSearch leadership for prioritization of deliverables as well as release.- Contributed to the open-source community of OpenSearch as part of the integration.Optimized Range Query Feature:- Designed and developed an optimized range query feature for applying business rules to personalized recommendations.- Ensured negligible latency impact while generating recommendations, maintaining high performance and user experience.

- Designed and development of algorithms, data structures and tools required for RapiDex(TM) process using C++, TCL scripting.- RapiDex(TM) is a motherboard customization service which gives design engineers the ability to tailor the exact features and form factor they need in an embedded motherboard with faster turnaround times.

Designed and development of algorithms, data structures and tools required for RapiDex(TM) process using C++, TCL scripting.

- Designed and developed autonomic framework and tool for network as well as access control policy management like Firewall policies, XACML policies.- Developed customized prototype of search engine using lucene indexing and nutch crawler for identifying foot-prints of suspected malicious users in the specified web domain.- Platform/Languages/Tools: C++, Java, Lucene, Nutch, Linux, Nessus.

- Reduced the time by 15% per user than existing application by developing network planning system workflow for British Telecom using open source MVC framework called JSF and J2EE Technologies like JSP, Servlet, JMS, and JDBC.- Developed interfaces of network planning system with other enterprise applications in British Telecom, UK using Web Services and JMS.- Platform/Technologies: Java, JSF, J2EE technologies, SQL (Oracle), JUnit.