Paresha Uchdadiya

LLM-powered semantic search using RAG, ElasticSearch, and Vertex AI embeddings. Improved search relevance by 60% and reduced noise across high-volume retail systems.

Enterprise-grade LLM platform for clinical decision support using agentic AI, RAG, and multimodal reasoning across EHR data. Improved diagnostic accuracy by 32%.

OCR + NLP pipeline for multilingual document processing using OpenCV, Tesseract, and PyMuPDF. Achieved 92% OCR accuracy and automated 30% of support queries.

Rasa-based NLU/Core chatbot with REST API integration for automated customer support. Reduced escalations by 18% through smart intent classification.

Scalable event-driven ETL pipeline for ingesting, transforming, and loading large-scale structured and unstructured data into vector and relational stores. Built with Kafka, Airflow, Spark, and Pinecone — enabling downstream AI pipeline consumption with automated retraining triggers.