Sai Pranav

At NVIDIA, I architected Retrieval-Augmented Generation infrastructure for internal LLM applications, building scalable retrieval pipelines across millions of documents. I developed Python ingestion workflows for extraction, cleaning, chunking, metadata enrichment, embeddings, and vector indexing.

I improved LLM response quality through embedding evaluation, reranking, hybrid retrieval, prompt experiments, and automated frameworks for retrieval accuracy, latency, citation quality, and answer faithfulness. I also built reusable SDKs, modular APIs, monitoring dashboards, and telemetry pipelines for production generative AI services.

Previously at Samsung Electronics America, I led machine learning solutions for power-demand forecasting and smart-grid optimization. I built end-to-end predictive modeling and ETL pipelines using Python and SQL, including Gradient Boosting, Random Forest, XGBoost, LightGBM, feature engineering, validation, retraining, and production monitoring.

I've also led multidisciplinary research software teams at the University of Rochester and conducted computer vision and medical-image-processing research at AT&T. I enjoy translating research and operational challenges into reliable AI products with engineering, research, and infrastructure teams.