Jayashree M Gadwal

AI Developer

Eximietas Design

Bengaluru

04.2024 - Current

Developed a resume parsing pipeline using Google Document AI to extract structured data, followed by custom post-processing to clean, normalize, and align outputs with client-specific requirements.
Verified extracted content through OpenCV-based bounding box visualization, ensuring accuracy and quality control in parsed results.
Built and integrated Retrieval-Augmented Generation (RAG) and Hybrid RAG systems using internal and external knowledge sources for enhanced LLM response generation.
Designed a modular agentic workflow with 4 query-resolution tools (Wikipedia, Google Search, RAG, Hybrid RAG); integrated into a voice-enabled system for dynamic, context-aware module triggering.
Engineered the real-time ASR module using FasterWhisper, incorporating wake word detection, speaker diarization, and query preprocessing for seamless voice interactions.
Developed a full-fledged Streamlit-based UI, integrating all modules into an interactive application for testing, demonstration, and internal deployment.
Built a video retrieval-augmented generation (Video RAG) system that processes uploaded videos by extracting frames and audio, chunking content, and generating summaries for each clip.
Deployed tools on a server-client architecture where an agent on the client side uses LangChain’s Model Context Protocol (MCP) to locally run components, coordinating with a master agent that dynamically selects and triggers the appropriate tools based on user queries
Enabled intelligent query routing and modular processing by integrating frame extraction, audio analysis, and summarization pipelines with agent-based orchestration.