Hi, I'm N.V.RaghuCharan

• Built a domain-specific RAG chatbot using OpenAI, ChromaDB, and Vanna, reducing ticket resolution time by 40%
• Developed fleet scheduling software with analytics dashboards, improving decision-making speed by 30%
• Implemented real-time data pipelines for operational insights and predictive maintenance

• Implemented real-time facial recognition attendance using OpenCV and InsightFace with 95%+ precision
• Integrated secure cloud pipelines for trip-level attendance tracking across multiple client deployments
• Optimized computer vision models for edge deployment, reducing inference latency by 35%

• Designed a CNN achieving 96% accuracy on MNIST using IBM AI Hardware Kit (aihwkit)
• Benchmarked neuromorphic inference, reducing computation by ~15% through analog computing
• Researched energy-efficient neural network architectures for edge AI applications

RAG Agent Chatbot

Domain-specific chatbot leveraging structured databases and large language models for intelligent query resolution and automated support.

• Implemented using Vanna, ChromaDB, OpenAI, and FastAPI
• Reduced query latency by 20% via optimized embeddings
• Achieved 40% improvement in ticket resolution time

EV Bus Scheduling Optimization

Constraint-based optimization engine for electric bus fleet operations, maximizing efficiency while considering battery constraints and route planning.

• Built with Google OR-Tools and Python
• Constraint-based scheduling with HTML & JSON outputs
• Improved operational efficiency by 30%

Synthetic Image Generation (GANs)

Data augmentation pipeline using Generative Adversarial Networks for low-data computer vision tasks, improving model generalization.

• Implemented with PyTorch and custom GAN architectures
• Improved generalization accuracy by ~10%
• Generated high-quality synthetic training data

Real-Time Facial Recognition System

Production-grade facial recognition system for automated attendance tracking with high accuracy and low latency performance.

• Achieved 95%+ precision using InsightFace and OpenCV
• Deployed on edge devices with optimized inference
• Integrated with cloud pipelines for scalability