Allen Gao

Built and shipped a production voice AI agent for enterprise healthcare with streaming speech-to-text, LLM, and text-to-speech pipeline, raising inbound call automation to 40%+. Designed evaluation harness for automated regression testing and built scalable agent infrastructure with full observability.

Engineered an LLM-based document audit platform processing ~1% of all U.S. patient medical records. Built generalized model inference pipeline on vLLM extracting structured data from clinical records at 99.99%+ accuracy.

Selected as 1 of 10 global teams for Amazon's Alexa Prize TaskBot Challenge 2; built LLM-powered voice-first multimodal task assistant deployed to 50M+ Alexa users. Built scalable model inference backend cutting operational latency by 50%.

Built video/audio media processing pipeline on AWS with OOP and multithreading, and a vector-database video query system improving search relevance by 55%. Engineered financial chatbot using LangGraph and GPT models.

Built SmartOCR, an end-to-end data collection and extraction pipeline for high school transcripts using AWS Textract, S3, and spaCy NER, achieving 95%+ accuracy on course, grade, and credit extraction.

Integrated AWS, GCP, and Tesseract OCR services into existing document workflows and ran field evaluations, improving OCR accuracy by 10%.