• Architected custom graph-based Python AI agent framework for multi-agent systems with configurable model switching, LLM tool calling abstraction, and execution tracing for agent workflow observability and debugging.

• Created competitor analysis website used daily by 50+ marketing teams by building Python/FastAPI backend with Redis caching to serve ad copies/images, and React/JS dashboard with KPI cards, charts, and ad copy previews

• Scraped 100+ competitor sites and Facebook Ad Libraries daily using automated Python/Selenium scraper with anti-bot bypass techniques including proxy rotation; deployed on GCE; stored ad text in Firestore and assets in GCS

• Built sub-1 second latency AI voice agent for natural phone conversations with interruptable responses and noise-resistant voice activity detection, using Python, Twilio, OpenAI, Docker, and GCP Cloud Run

• Shipped 50+ features to 10,000 users within a production lecture video platform by taking leadership role in developing new assistive technologies to improve user experience for persons with disabilities, using React/Redux/JS.

• Designed feature to create digital notes from lecture videos with real-time editing and previewing that mirrors final output across ePub, PDF, HTML, and LaTeX exports, using React, Redux, and JavaScript

• Improved accessibility for blind students by creating audio description feature to generate, edit, and speak image descriptions within video player with adjustable text-to-speech parameters and screen reader compatibility

• Co-authored five peer reviewed papers assessing the needs of undergraduate students in engineering classes and how to use new assistive technologies to improve education for students with disabilities.

• Designed surveys and conducted interviews to assess the opinions and needs of students and instructors in undergraduate courses.

• Designed training course to teach faculty best practices in making courses accessible for all students using Universal Design for Learning (UDL) principles.

• Assisted over 400 students with their group work during discussion sections.

• Reviewed and wrote homework and exam questions.

• Graded and provided meaningful feedback on students' code, homework, and exams.