Mountain View, California, United States
• Contributed to developing a full-stack web application for a confidential AI-powered platform leveraging generative AI, utilizing React and Redux for micro frontend architecture and Java for backend services. • Built personalized features, including profile management, interest-based content delivery, and location-aware experiences, to enhance user engagement. • Engineered secure, RESTful APIs to manage sensitive user data, ensuring robust compliance with privacy regulations and industry standards. • Led product launch efforts by conducting load tests to identify system breakpoints and uncover vulnerabilities while building monitors and dashboards to track system performance and ensure long-term reliability
• Independently developed and delivered geolocation-based search capabilities for BMW's Alexa Custom Assistant, driving seamless interactions across multiple i18n locales and enhancing the user experience across regions. • Collaborated across teams to deliver the Point of Interest (POI) Calling feature across 14 locales for BMW vehicles, enabling region-specific customization. • Designed and implemented a Prompt System for BMW’s offline voice assistant, enabling dynamic prompt generation based on navigation and Point of Interest (POI) data, to provide localized and context-aware responses. • Developed end-to-end testing infrastructure for Alexa Auto’s local search features, ensuring more reliable feature releases and improved testing accuracy. • Provided operational support for Alexa Local Information features by resolving long-standing, over-SLA critical bugs and security issues.
• Worked on the project SEARCCH that is a community-driven platform to aid researchers to share and reuse experimental artifacts in the cybersecurity domain and the project is supported by the National Science Foundation. • Automated the process of collecting data and files related to artifacts from the URL provided by the user. This data is further processed and stored in the database using sqlalchemy. Reduced the need of manual data entry for 85 percent of artifacts. • Constructed a markdown parser to convert markdown text to an abstract syntax tree and parse it to collect data that was missed initially during data extraction. Improved Metadata for 20 percent of all artifacts and 67 percent of GitHub artifacts. • Keywords improve the quality of search so identifying relevant keywords automatically was necessary for the hub. • Built a keyword extractor to get keywords automatically from the publication pdf of the experimental artifacts using Text Rank Algorithm thus adding meaningful keywords as metadata for nearly 70 percent of the artifacts. • Improvised keyword extraction for 60 percent of code artifacts by extracting comments, names of class, methods, and attributes from the source code files and processing further depending on the parts of speech of these words.
Teaching Assistant for Computer Organization.