At TINY ROCKET, I developed a TypeScript hallucination-detection system and refactored the LLM evaluation system to support OpenAI, Anthropic, Gemini, and Azure OpenAI. I also built an internal MCP-based agent that processed more than 70 meeting transcripts, summarized them, and created and assigned tasks in Notion.
At Edubaloo, I built a hybrid RAG chatbot serving more than 34,000 students, achieving a 91% retrieval hit rate and 88% answer relevance in an evaluation of 500 questions. I’ve also developed data-processing pipelines and distributed systems, including a personal price aggregation project using Node.js, Express, Redis, and BullMQ.
