Projects
HeritXR – RAG-Based Heritage Question Answering System
AI Research | Generative AI | Retrieval-Augmented Generation
Developed a Retrieval-Augmented Generation (RAG) framework to provide historically validated and context-aware answers within an immersive Virtual Reality heritage environment. The system retrieves information from verified heritage sources before generating responses, ensuring accurate, explainable, and trustworthy visitor guidance.
Key Project Highlights:
Validated heritage knowledge retrieval using vector databases
Semantic search for accurate information discovery
Retrieval-Augmented Generation (RAG) to reduce AI hallucinations
Context-aware conversational responses
Source-grounded answers with full traceability
Real-time integration with Virtual Reality environments
Speech-enabled AI assistant for interactive guidance
FastAPI and LangChain-powered backend architecture
Technologies & Tools:
OpenAI, Gemini, LangChain, Ollama, FastAPI, Node.js, MySQL, Vector Database, Semantic Search, Prompt Engineering, Conversational AI, JSON, REST APIs, Retrieval-Augmented Generation (RAG)
Project Outcomes:
95% Knowledge Retrieval Accuracy
100% Source Traceability
Reduced Hallucination Rate
Real-Time Heritage Question Answering
Trusted and Explainable AI Responses
HeritXR – Visitor Persona Identification Using Agent Debating
Artificial Intelligence | Multi-Agent Systems | Personalization
Developed an intelligent visitor persona identification framework that utilizes Agent Debating to analyze visitor interactions and determine individual interests within a virtual heritage environment. Multiple AI agents collaboratively evaluate behavioral patterns and preferences to accurately classify visitor personas, enabling personalized storytelling, recommendations, and adaptive tour experiences.
Key Project Highlights:
Multi-Agent AI architecture for persona identification
Agent Debating mechanism for collaborative decision making
Dynamic visitor interest profiling
Real-time behavioral analysis
Personalized recommendation generation
Adaptive storytelling based on visitor preferences
Context-aware user modeling
High-accuracy persona classification system
Technologies & Tools:
Artificial Intelligence, Multi-Agent Systems, Agent Debating Framework, Machine Learning, Visitor Profiling, Recommendation Systems, Python, FastAPI, JSON, Behavioral Analytics, Personalization Engine
Project Outcomes:
95% Persona Identification Accuracy
Improved personalization of heritage experiences
Enhanced visitor engagement through adaptive content delivery
Real-time user preference analysis
Intelligent recommendation generation based on visitor interests
Fungal Colony Detection in Yoghurt Using Image Processing
Developed an image processing solution to detect and analyze fungal contamination in yoghurt samples. The system applies computer vision techniques for image enhancement, segmentation, and feature extraction to support early identification of microbial growth and improve food quality assessment.
Key Project Highlights:
Image preprocessing and enhancement
Fungal colony detection and segmentation
Feature extraction from contaminated samples
Automated image analysis workflow
Food quality assessment support
Python-based computer vision implementation
Technologies & Tools:
Python, OpenCV, NumPy, Image Processing, Computer Vision
Serene Minds – Mental Wellness Website
Developed a responsive website focused on promoting mental well-being and emotional wellness. The platform provides users with accessible resources, informative content, and a calming user experience designed to support mental health awareness and personal growth.
Technologies & Tools:
HTML, CSS, JavaScript, Responsive Web Design
Project Link: