Gauri Godghase

I've built end-to-end agentic AI and LLM systems for Opelsoft's client, Alo Yoga, turning 170,000+ support conversations into a structured root-cause taxonomy that replaced monthly manual review.

I developed a real-time voice AI agent using LiveKit and the OpenAI Realtime API, connecting retrieval, tool calls, and live order lookups across a 348-article knowledge base. I also reduced analytics-query latency from roughly 30 seconds to 5 seconds and productionized full-population LLM runs on AWS Batch and OpenAI APIs.

At Workiva, I built a generative AI research assistant with FastAPI, GPT, LLaMA, RAG, AWS Bedrock, OpenSearch, and multi-agent workflows. My work evaluated quality, latency, reliability, and cost to guide production architecture decisions.

As a Research Engineer at San Jose State University, I studied human- versus LLM-generated text across 1.5M+ samples using BERT, RoBERTa, LSTM, GloVe, and Word2Vec. This work resulted in arXiv and Springer publications, grounded in controlled experiments, statistical evaluation, and rigorous error analysis.