Full-Stack AI Developer at Upwork

AI Conversational Admissions Assistant An end-to-end conversational AI system that helps university staff and students get quick, accurate answers. ● Designed and implemented an intelligent query processing pipeline with LLM-based query reconstruction (pronoun resolution, context merging), domain classification, and hybrid retrieval (vector embeddings + BM25) with metadata filtering to deliver precise, contextually relevant responses. ● Built an automated evaluation system for search quality and answer validation, achieving 88.2% retrieval accuracy on production FAQ data with integration tests for continuous quality assurance. ● Developed end-to-end chat system with SSE streaming, Redis-backed session management, structured logging (structlog), and graceful error handling with fallback mechanisms to ensure reliability. ● Integrated ElevenLabs conversational AI (TTS/ASR) with custom audio streaming endpoints and real-time voice interaction capabilities.