Projects

Medvision is a full-stack web application that allows users to upload medical images and get a diagnosis from an AI model.

• I built a deep learning model to classify medical images (e.g., pneumonia detection from X-rays). I applied OpenCV-based preprocessing and data augmentation to improve accuracy. I developed a full-stack web app with Flask (backend) and React (frontend), enabling real-time image uploads and AI based classification.

Zaruram is a SaaS platform that helps businesses manage their inventory, clientbase, operations, and improve their efficiency.

• I developed an AI-assisted business management SaaS platform, integrating Supabase for authentication and databasemanagement, with OAuth and JWT for secure user access. I implemented automated document handling and analytics, leveraging TypeScript and AI-powered insights to streamline business operations and Optimized cloud storage and data retrieval, utilizing AWS S3 for scalable file management and enhancing system efficiency.

Built a machine learning system for real-time anomaly detection in network traffic, leveraging 5GB+ log data and time-series features for accuracy.

• Developed a hybrid model combining statistical methods (Isolation Forest, PCA) and deep learning (LSTMs, Autoencoders), achieving 93% detection accuracy.
• Designed an automated pipeline for data ingestion, preprocessing, model training, and inference, reducing manual effort by 80%.

Developed a network intrusion detection system (IDS) to monitor and analyze packet data for malicious activities.

• Implemented deep packet inspection and anomaly detection using Scapy and machine learning algorithms.
• Integrated automated alerting mechanisms via email and Slack for real-time threat notifications.

Poisha is a full-stack web application designed to provide users with personalized flashcards for efficient learning. It adapts to the user’s progress to ensure a tailored experience.

• Developed the frontend using React.js with a modular component architecture.
• Implemented user authentication and privacy features with Firebase Realtime Database.
• Created dynamic, responsive interfaces optimized for various devices.

EcoLar is a mobile app built with React Native that promotes eco-friendly transportation choices by calculating carbon footprints and tracking user achievements.

• Implemented a carbon footprint algorithm that compares various transportation methods.
• Developed a gamification system with MongoDB to incentivize eco-friendly choices.
• Utilized REST APIs to fetch real-time data for public transportation and traffic.

Zmaka is a weather prediction application that helps farmers make informed decisions about crop selection based on seasonal data.

• Led the development lifecycle from concept to deployment.
• Built machine learning models with TensorFlow to predict agricultural yields based on weather patterns.
• Developed a user-friendly web interface with React.js to visualize weather forecasts and predictions.

ChatCore is a lightweight web server built from scratch using C, designed for real-time messaging and user connections.

• Developed custom TCP/IP socket programming for real-time messaging.
• Built a persistent storage system with optimized file I/O for fast read/write operations.
• Ensured server stability and performance under concurrent user requests by implementing error handling and memory management.