AI Research Engineer at WESEE – Naval R&D Lab

Led the development of an offline, self-hosted web application using LLaMA 3.1 8B Instruct, integrating a LangChain-based RAG pipeline over a classified naval documents knowledge base. Designed a FastAPI-based backend with a ReactJS frontend and deployed in an air-gapped, CUDA-enabled environment using vLLM for low-latency, memory-efficient inference. Architected data ingestion, indexing, and retrieval workflows with LlamaIndex and VectorDBs to ensure secure access. Implemented security hardening and access controls suitable for defense-grade operations and containerized the stack for reproducible deployments across isolated nodes. Built an LLM-powered document-to-audio pipeline using Coqui TTS, generating Q&A, summaries, objective questions, and FAQs with LLaMA. Developed a satellite image analytics system using YOLOv8 with Oriented Bounding Boxes for maritime zone object detection, including tiling, normalization, and resolution scaling to optimize inference and compute real-world di

AI Intern at Object Automation System Solutions Pvt. Ltd.

Contributed to a collaborative filtering-based movie recommendation system using Flask. Hosted and administered a Learning Management System on a Linux-based VPS, gaining experience with deploying and maintaining web services in a server environment.

AI Research Engineer at WESEE – Naval R&D Lab

Led development of an offline, self-hosted web application using LLaMA 3.1 8B Instruct, integrating a LangChain-based RAG pipeline over a classified naval documents knowledge base. Designed a FastAPI backend with a React frontend and deployed in an air-gapped, CUDA-enabled environment using vLLM for low-latency, memory-efficient inference. Architected an LLM-powered document-to-audio pipeline for podcast-style narrations using Coqui TTS, generating Q&As, summaries, objective questions, and FAQs. Implemented a satellite image analytics system using YOLOv8 with Oriented Bounding Boxes for maritime object detection.