Harshitha Mittapalli

Developed and fine-tuned Retrieval-Augmented Generation (RAG) workflows by integrating vector stores and embedding models, achieving an 18% boost in response latency and contextual accuracy. Designed and deployed scalable ChromaDB based vector stores to index high-dimensional text data, resulting in a 25% uplift in semantic retrieval efficiency. Automated data chunking and embedding generation pipelines using LangChain and Google Gemini LLM, transforming large datasets into vectorized formats for fast, context-aware retrieval. Assembled and evaluated training datasets to improve embedding fidelity, leading to a 12% increase in retrieval precision and enhancing real-time AI interaction scalability.

Developed and refined Retrieval-Augmented Generation (RAG) workflows by integrating vector stores (ChromaDB) and embedding models, improving contextual accuracy and response latency by 18%. Designed scalable, ChromaDB-based vector stores to index high-dimensional text data, boosting semantic retrieval efficiency by 25%. Automated data chunking and embedding generation pipelines using LangChain and Google Gemini LLM, transforming large datasets into vectorized formats for fast, context-aware retrieval. Assembled and evaluated training datasets to improve embedding fidelity, contributing to a 12% increase in retrieval precision and enabling scalable real-time AI interaction.

Engineered backend services for a tax filing web application using Java, Spring Boot, and MySQL, applying Object-Oriented Design principles for modular, secure, and scalable API architecture. Boosted system responsiveness and deployment speed by optimizing front-end state handling, backend services, CI/CD pipelines in Azure, and Agile sprint processes. Integrated ReactJS and TypeScript front-end features such as debounced inputs, micro frontends, real-time validations, improving data input efficiency by 10%. Enhanced UI responsiveness with modular, interactive components, using Zustand for efficient state management to deliver a user-friendly digital tax filing experience. Built a Java Spring Boot microservice using OOD principles and Azure Document Intelligence for OCR to extract text from 200+ PDFs/images and convert into structured JSON. Designed end-to-end ETL pipelines with Azure Data Factory, transforming raw client data into clean MySQL structured datasets and reducing processin