MR.

Projects

Here are some of the projects I have worked on.

01

AI News Aggregator

AI-Powered News Aggregation System

An intelligent, AI-powered news aggregation system using GPT-4.1 and GPT-4o-mini that automatically scrapes, processes, and delivers personalized AI news digests via email.

PythonOpenAI GPT-4.1PostgreSQLLangChainDocker
  • ‣Built multi-source content scrapers for YouTube (with transcript extraction), OpenAI news, and Anthropic blog posts
  • ‣Implemented AI-powered summarization using GPT-4o-mini and personalized article ranking using GPT-4.1
  • ‣Designed a curator agent that scores articles (0-10) based on user profile relevance
  • ‣Developed automated email digest pipeline with beautifully formatted HTML newsletters via Gmail SMTP
  • ‣Managed persistent storage with PostgreSQL and SQLAlchemy with duplicate detection and incremental processing
02
Live

Needflicks

AI-Powered Movie Streaming Platform

A full-stack movie streaming platform with AI-powered sentiment analysis using LangChain & OpenAI, personalized recommendations, and JWT-based secure authentication.

GoReactMongoDBOpenAILangChain
  • ‣Built a Go/Gin backend with MongoDB, featuring JWT authentication with access/refresh token rotation via HTTP-only cookies
  • ‣Integrated LangChain with OpenAI for automated sentiment analysis of admin movie reviews, auto-assigning rankings
  • ‣Developed a personalized recommendation engine based on user favorite genres and AI-classified movie rankings
  • ‣Created a React 19 frontend with role-based access control (Admin/User), video streaming via React Player, and responsive Netflix-inspired UI
  • ‣Implemented secure architecture with bcrypt password hashing, CORS middleware, input validation, and admin-only routes
03
Live

Crown Clothing

E-commerce Application

A modern React & Vite-based e-commerce platform for browsing products, managing carts, and secure checkout with Stripe.

ReactViteFirebaseFirestoreStripe
  • ‣Implemented user authentication using Firebase Auth (Google Popup & Email/Password)
  • ‣Fetched and stored products, categories, and user data in Firestore
  • ‣Managed global state for cart and shop with React Context API
  • ‣Integrated Stripe payments via Netlify serverless function for secure checkout
  • ‣Styled with Sass and styled-components for responsive UI
04

Go E-commerce Platform

Fullstack Ecommerce Web Application in Go

Go Ecommerce Web Application with MongoDB, user authentication & go modules for dependency management

GolangMongoDBDocker
  • ‣Implemented secure user authentication and authorization system
  • ‣Built RESTful APIs for order management and inventory control
  • ‣Integrated payment processing system with error handling
  • ‣Developed real-time inventory tracking system
  • ‣Created admin dashboard for order and inventory management
05
Live

YouTube Clone

Full-Stack Video Platform

A full-featured YouTube clone built with Django and ImageKit.io for video storage and HLS adaptive bitrate streaming with user authentication and channel pages.

DjangoPythonImageKit.ioSQLiteRailway
  • ‣Built a Django backend with user authentication (register, login, logout) and video upload support up to 100MB (MP4, WebM, MOV, AVI)
  • ‣Integrated ImageKit.io for video storage with HLS adaptive bitrate streaming from 240p to 1080p
  • ‣Implemented like/dislike voting system with real-time updates and channel pages for user video collections
  • ‣Developed automatic username watermarking on thumbnails and custom thumbnail upload functionality
  • ‣Designed a responsive dark-theme UI inspired by YouTube with WhiteNoise static file serving and Railway deployment
06

Toll Calculation using Go Microservices

Golang Microservices

A suite of Go-based microservices processing simulated OBU data to calculate toll fees, leveraging Kafka, gRPC, and Prometheus.

GoKafkagRPCPrometheusGrafana
  • ‣Built four microservices (OBU, Data Receiver, Distance Calculator, Invoice Aggregator) in Go
  • ‣Orchestrated real-time data flow with Apache Kafka messaging
  • ‣Exposed REST & gRPC endpoints, including a go-kit based aggregator
  • ‣Instrumented with Prometheus & Grafana for metrics and monitoring
  • ‣Containerized and orchestrated via Docker Compose
07

Go URL Shortener

URL Shortener Service

High-performance URL shortening service built with Go and Redis

GoRedis
  • ‣Built scalable URL shortening service using Go-Fiber framework
  • ‣Implemented Redis for high-speed URL mapping storage
  • ‣Developed rate limiting and security features
  • ‣Created analytics dashboard for URL tracking
  • ‣Achieved sub-millisecond response times
08

JSON Object Indexing Platform

Health Insurance Data Indexing System

A high-performance indexing system for healthcare data processing

PythonFlaskElasticsearchRedisRabbitMQ
  • ‣Developed a scalable Python-Flask API that processes and indexes complex healthcare JSON documents
  • ‣Implemented real-time data processing using RabbitMQ message queues with 99.9% reliability
  • ‣Optimized search performance using Elasticsearch, achieving 50ms average query response time
  • ‣Integrated Redis caching layer reducing repeated query load by 40%
  • ‣Designed fault-tolerant architecture handling 100K+ daily insurance claim records
09

ETL Data Pipeline

NYC Ride Data Pipeline

Scalable ETL pipeline for processing NYC ride data

PythonPandasDaskDuckDBDocker
  • ‣Developed high-throughput data processing pipeline
  • ‣Implemented parallel processing using Dask
  • ‣Optimized data storage using DuckDB
  • ‣Created automated data quality checks
  • ‣Processed over 1TB of ride data efficiently
10

ML-Powered Credit Risk Analysis

Credit Limit Allocation & Loan Defaulter Prediction

Advanced machine learning system for credit risk assessment and loan default prediction

PythonScikit-learnPandasNumpySeaborn
  • ‣Developed ML models achieving 92% accuracy in default prediction
  • ‣Processed and analyzed 100K+ customer records for training
  • ‣Implemented feature engineering for 20+ financial indicators
  • ‣Created interactive dashboards for risk visualization
  • ‣Reduced false positives by 35% compared to traditional methods
11

Graph-based RAG System

CVE-GraphRAG using Kafka and Neo4j

Advanced CVE analysis system using graph databases and RAG architecture

KafkaOpenAIStreamlitSnowflakeNeo4j
  • ‣Built scalable data pipeline processing 100K+ CVE records
  • ‣Implemented real-time vulnerability detection system
  • ‣Developed graph-based query optimization reducing response time by 60%
  • ‣Integrated OpenAI API for intelligent vulnerability analysis
  • ‣Created interactive visualization dashboard for security insights
12

CVE-Guardian System

CVE-Guardian and Container Image Scanning

Advanced vulnerability detection system using Llama-Index and Langchain

Kafka in GoOpenAI APITailwindCSSPGVector
  • ‣Developed an automated vulnerability scanning system for container images
  • ‣Integrated Llama-Index and Langchain for intelligent vulnerability analysis
  • ‣Implemented real-time monitoring and alert system
  • ‣Created comprehensive reporting dashboard
  • ‣Reduced false positive detection rate by 45%
13

Depth Anything V2 Vision System

Computer Vision Project

Comprehensive real-time image and video processing toolkit using OpenCV and neural networks

C++OpenCVONNX RuntimeCMake
  • ‣Developed advanced C++ applications for real-time image and video processing
  • ‣Integrated Depth Anything V2 neural network for accurate depth estimation
  • ‣Implemented multiple custom filters including optimized Gaussian blur and Sobel edge detection
  • ‣Created interactive face detection with selective colorization features
  • ‣Built depth-based visual effects including fog simulation and embossing
14

SiglipVision Transformer

Computer Vision & Deep Learning

From-scratch PyTorch implementation of Google's SiglipVision model with validation against Hugging Face

PyTorchHuggingFacePythontorchvision
  • ‣Built complete implementation of Google's vision transformer architecture for image understanding
  • ‣Developed custom vision embeddings layer with patch extraction and positional encoding
  • ‣Implemented multi-head attention mechanism with scaled dot-product attention
  • ‣Created validation system to verify output compatibility with Hugging Face transformers library
  • ‣Designed visualization tools for patch embeddings and model attention patterns
15

2D Object Recognition System

Real-Time Computer Vision

A comprehensive C++ real-time 2D object recognition system using custom-built computer vision algorithms for translation, scale, and rotation invariant object identification.

C++OpenCVCMake
  • ‣Implemented K-means (ISODATA) adaptive thresholding, morphological operations, and two-pass connected components analysis from scratch in C++
  • ‣Engineered an 11-feature vector including Hu moments, compactness, eccentricity, and aspect ratio for robust shape description
  • ‣Built Nearest Neighbor and K-Nearest Neighbor classifiers with multiple distance metrics (Euclidean, Cosine, L1, L-infinity)
  • ‣Developed an interactive training mode with CSV-based object database and confusion matrix evaluation framework
  • ‣Achieved 85-95% accuracy at 30+ FPS with real-time oriented bounding box and axis visualization
16

Multi-Pattern Weather Forecasting System

Object-Oriented Design & Architecture

Sophisticated Java application demonstrating seven design patterns through a weather forecasting system

JavaOO DesignDesign Patterns
  • ‣Designed a system implementing Builder, Factory, Singleton, Decorator, Strategy, Composite, and Adapter patterns
  • ‣Developed multi-location weather forecasting with temperature and precipitation modeling
  • ‣Created regional weather analysis through composite location objects
  • ‣Implemented multiple algorithmic weather modeling strategies
  • ‣Built an extensible architecture allowing easy addition of new locations and prediction models
Good code