Featured Work & Projects
A curated selection of my professional work and personal projects, showcasing expertise in GenAI & Agentic AI, Computer Vision, and Full-Stack Development.
OnePiece AI Agent System
Sophisticated full-stack autonomous AI agent platform designed to orchestrate specialized AI agents for complex task execution. Leverages LangGraph to manage agent state and collaboration, allowing multiple specialized agents (e.g., search, analysis, execution) to work together to solve multi-step problems. Features a real-time streaming interface, high-performance caching, and a comprehensive mission dashboard for human-in-the-loop monitoring.
Real-Time Threat Detection Model
A real-time intelligent surveillance system capable of detecting weapons, intruders, and suspicious objects using deep learning. Optimized GPU-based inference achieving 92% mAP@50 with sub–100ms latency for live CCTV streams.
Enterprise RAG Knowledge Assistant
Production-ready Retrieval-Augmented Generation (RAG) platform designed for enterprise-grade document search and AI-powered Q&A. This system allows users to upload complex PDF documents and interact with them through a natural language interface. Features a high-performance ingestion pipeline that parses, chunks, and indexes documents into a vector database (FAISS). Implements semantic search to retrieve contextually relevant information and uses an LLM (Mistral-7B) to generate precise answers with inline citations and real-time performance metrics (latency, similarity scores, and chunk retrieval counts).
Facial Recognition System
Secure facial recognition system using ArcFace embeddings and cosine similarity matching. Implements blink-based liveness detection to prevent spoofing and unauthorized access, with MySQL logging and Streamlit dashboard.
NeuroAI - Brain Tumor Analysis & 3D Visualization Platform
Comprehensive medical imaging system designed to assist clinicians in the detection, segmentation, and classification of brain tumors from multi-modal MRI scans (T1, T1ce, T2, FLAIR). Features a multi-stage AI pipeline integrating a residual DnCNN-style network for denoising, an Attention UNet++ for pixel-level tumor segmentation, and a transfer-learned ResNet50 for classification. Generates 3D meshes (OBJ) from NIfTI volumes for spatial mapping and uses Grad-CAM++ for explainable AI (XAI) heatmaps. Includes automated PDF reporting and an interactive 3D dashboard.
Lung Nodule Detection
Automated lung nodule detection system using 3D CT scan analysis. Employs a custom CNN architecture with false-positive reduction and volumetric segmentation to assist radiologists in early-stage pulmonary cancer screening.
API Security Gateway
High-performance, production-ready API Security Gateway built with Node.js. Acts as a centralized security layer that sits in front of backend services, enforcing critical security policies, rate limiting (Token Bucket via Redis), and monitoring before requests reach upstream servers. Supports dual-layer authentication (JWT and API Keys), Prometheus integration for real-time observability, distributed tracing (X-Request-ID), and structured logging with Pino.
Smart Agri Advisor
AI-powered precision agriculture advisory system providing real-time crop health monitoring, soil analysis recommendations, and yield prediction. Integrates sensor data streams with an LLM-driven advisory chatbot for farmers.
Interview RAG Bot
AI-powered interview preparation assistant built on a RAG architecture. Ingests domain-specific study material and generates tailored Q&A sessions, evaluates answers with rubric scoring, and provides contextual feedback using LLMs.