Projects
AI Secretary
Built an LLM secretary that can hold conversation with an user and respond with owner’s voice.
- Used Twilio integration for incoming calls and streaming response audio, configured OpenAI to ElevenLabs streaming pipeline for text generation and voice cloning. Used LangChain to provide in-conversation memory to agents.
- Designed agent architecture to stream data with websockets and bidirectional communication to optimize latency
AI Rate My Professor
RAG chat app to find professors and ratings
- Created a RAG (Retrieval Augmented Generation) chat app for community college students to find professors and ratings
- Used OpenAI API for response generation, setup Pinecone database for knowledge-base search, and Next.js + React for frontend
Flashcard SaaS
AI flashcard generator platform.
- Used OpenAI API for flashcard generation and integrated a paywall with Stripe API.
- Implemented frontend and dynamic routing with Next.js + React, hosted backend on Firebase, and deployed with Vercel
DeepSafe
Deepfake detection for video conferencing platforms.
- Selected from a highly competitive pool of 50+ teams to participate in a university-wide competition, and pitched to Berkeley SkyDeck incubator program.
- Refined MVP based on user needs by conducting usability tests with 20 potential users and building high-fidelity mockups using Figma.
