Projects

I build practical applications that connect AI, data, and software development. My projects span explainable machine learning, restaurant analytics, healthcare research prototypes, and web and mobile applications, turning technical ideas into tools for learning, analysis, and everyday workflows.

Selected Kaggle Projects

A selection of my public notebooks exploring language, explainable AI, scalable analytics, and financial crime detection. Explore the code, methods, and experiments on Kaggle.

Bangla NLP

Ensemble-Based Multi-Label Bangla Cyberbullying

Detects overlapping categories of harmful Bangla text using word and character features, weighted soft voting, and label-specific thresholds.

TF-IDF · Ensemble learning

View notebook
Cybersecurity

Internet Firewall Action Classification

Classifies firewall actions from network traffic, combining feature engineering and a neural network with SHAP explanations of model predictions.

Deep learning · SHAP

View notebook
Big Data Analytics

Scalable Used-Car Price Prediction with PySpark

Builds a scalable pricing workflow with data cleaning, feature engineering, model comparison, and residual analysis for used-car listings.

PySpark · Regression

View notebook
Financial Crime

Credit Card Fraud Transaction Detection

Explores transaction fraud detection through classifier comparisons, resampling experiments, and a soft-voting ensemble.

Classification · Ensemble learning

View notebook
Financial Analytics

Credit-Risk Classification with PySpark

Explores credit-risk classification using PySpark, applying distributed data processing to a financial prediction task.

PySpark · Classification

View notebook
Cybersecurity

IoT Network Anomaly Detection Using PySpark

Investigates unusual patterns in IoT network data using PySpark for a scalable network-security analysis workflow.

PySpark · Anomaly detection

View notebook
Computer Vision

Fish Species Detection

Explores fish species recognition from images, applying computer vision to distinguish species in a classification task.

Image classification · Species recognition

View notebook
Natural Language Processing

BNLP Research

Investigates Bangla natural language processing through a public research notebook focused on language-data experiments.

Bangla · Language research

View notebook
Explore all my Kaggle notebooks

An Explainable AI Decision-Support Framework for Service Delay Prediction in Multi-Restaurant Operations

Jan 2026 – Mar 2026

Predicts four restaurant service-delay outcomes using operational data, machine learning, SHAP explanations, and a recommendation layer.

Developed an explainable AI decision-support framework to predict and interpret operational delays across multi-restaurant operations. The project used transaction-level POS data from eight Brother Marcus restaurant locations, covering January 2025 to March 2026, including order, item, modifier, payment, and service-level information.

The system predicted four key operational outcomes: food service delay, drinks service delay, order completion delay, and table turnover delay. Multiple machine learning models were evaluated, including Logistic Regression, Random Forest, Gradient Boosting, Gaussian Naïve Bayes, and XGBoost, with XGBoost producing the strongest overall performance.

AI Agent for Brother Marcus Restaurants Using Agentic AI Workflow

Oct 2025 – Dec 2025

A multi-agent restaurant analytics chatbot using natural-language queries, a Gemini orchestrator, specialist agents, and controlled data APIs.

Designed and implemented a multi-agent AI chatbot for Brother Marcus Restaurants to support executive-level restaurant analytics using natural language queries. The system uses an agentic AI workflow with a Gemini-powered orchestrator, specialised agents for covers, operations, menu/sales, order insights, and automated email reporting, connected to Toast POS and Brother Marcus internal data through a controlled API layer.

OTA System using Sabre API

Aug 2023 – Nov 2023

A Sabre API travel system supporting flight information, fare revalidation, PNR creation, and booking workflows.

The OTA System simplifies the entire travel process, providing a user-friendly interface that allows travelers to explore, plan, and book their trips effortlessly. Powered by Sabre API, the system taps into a vast repository of real-time data, ensuring users can access the most accurate and up-to-date information on flights, hotels, car rentals, and more.

ChatAI: A chatbot that uses OpenAI’s GPT-3

Apr 2023 – May 2023

A GPT-3 conversational chatbot built with web technologies including JavaScript, Vite, and Node.js.

The primary objective of the ChatAI project is to create a highly interactive and intelligent chatbot that can engage in natural and meaningful conversations with users. Leveraging the power of OpenAI’s GPT-3 language model, the chatbot aims to provide a seamless and realistic conversational experience.

The project utilizes various web technologies such as HTML, CSS, JavaScript, Vite.js, and Node.js to develop a web-based platform for the chatbot. The web interface ensures easy accessibility for users across different devices and platforms. By making the chatbot available online, it allows users to engage with the system from anywhere with an internet connection.

GenImage: An OpenAI Image Generation App using MERN Architecture

Apr 2023 – May 2023

A text-to-image application combining OpenAI APIs with React, MongoDB, Express, Node.js, and Cloudinary.

The web app we have developed is a powerful tool that leverages OpenAI’s state-of-the-art GPT-3 text-to-image generation model to create stunning and realistic images based on user input. With a seamless integration of cutting-edge technologies such as Vite (React), MongoDB, Express, Node.js, Cloudinary, and OpenAI, this application provides an intuitive and user-friendly experience for generating images from text descriptions. To access the web app, simply visit the following URL: https://open-ai-image-generation-using-mern.vercel.app/

With this web app, users can experience the creative potential of AI and explore the fascinating world of text-to-image generation. Whether for artistic inspiration, design prototyping, or simply curiosity, our application empowers users to transform their textual ideas into stunning visual representations.

VidTube: A Video Streaming Website Built with React, Material UI, and RapidAPI

Apr 2023 – May 2023

A responsive video browsing and streaming interface built with React, Material UI, and RapidAPI.

Overview VidTube is an impressive video streaming website developed using React, Material UI, and RapidAPI. This platform offers users the ability to explore and watch videos from YouTubes extensive collection. By leveraging the power of React and Material UI, VidTube delivers an intuitive and visually appealing user interface, while RapidAPI facilitates seamless integration with YouTube API for data retrieval.

To experience VidTube firsthand, users can visit the live demo at https://cozy-starlight-5e2064.netlify.app/.

VidTube combines cutting-edge technologies and APIs to deliver an exceptional video streaming experience. With its intuitive interface, high-quality playback, and extensive features, VidTube is poised to captivate users and provide them with a platform to explore and enjoy YouTubes vast video library.

REST APIs, React.js and +1 skill

An Online Job Hiring Web Application

Jan 2023 – Mar 2023

Niyog is a recruitment application with CV creation, job search, applications, candidate search, and online interviews.

PawCenter: A Vet Hiring Pet Adoption and Foster Reservation System

Oct 2022 – Dec 2022

A veterinary services platform supporting pet adoption, foster reservations, user accounts, and reservation management.

Overview Pawcenter: is a Vet Hiring Pet Adoption and Foster Reservation System is a comprehensive web application catering to the needs of a veterinary clinic while providing a user-friendly platform for pet adoption and foster reservation. The application ensures a seamless experience for users by allowing them to create accounts and access essential features. Users can easily explore information about the veterinary clinic, including its location, working hours, and offered services.

UniTrack: A RFID Attendance System

Oct 2022 – Nov 2022

An RFID attendance system using Bootstrap, PHP, and MySQL to manage university courses, devices, teachers, and students.

Overview UniTrack is a comprehensive attendance management system designed specifically for universities. It has been developed using Bootstrap, PHP, MySQL, and RFID modules to provide a user-friendly interface and efficient administration. The system enables universities to seamlessly manage semesters, departments, RFID devices, courses, programs, teachers, and students. With UniTrack, administrators can easily create classes and assign teachers and students to those classes. Teachers, upon receiving their account credentials, can effortlessly add students to the classes assigned to them by the admin. The attendance tracking process is simplified as teachers can activate the RFID device in their class, and as students swipe their cards, attendance is automatically recorded. UniTrack simplifies attendance management for universities, providing accurate and convenient tracking while reducing administrative burdens.

EveCode: Learning Management System

May 2022 – Oct 2022

A learning management platform with a built-in compiler, coding courses, class management, and file handling.

EveCode is a comprehensive learning management system that includes an in-built compiler for popular programming languages such as C, C++, PHP, JavaScript, and Python. It provides a platform for users to write, run, and debug their code with ease. The system is designed to cater to both individuals looking to improve their coding skills as well as academic organizations that want to provide their students with a seamless learning experience.

Key Features In-built compiler system: With EveCode, you can write, debug and run your code directly from the platform without the need to switch to a separate environment. File management: Create, edit, delete, and download repos with ease. The system automatically overrides the previous code file each time you run the code. Free compiler: Even without logging into the system, you can use the free compiler to run your code. Your files will not be saved on the server in this case. Online courses: Sign up for courses designed and made by professional teachers and solve the designed lessons. Attempt the lessons directly on the compiler and view the prompt, write code and see the output in real-time. University classes: Join university classes if your academic organization purchases a subscription and uploads their course onto the system. Software used EveCode has been developed using a combination of the latest web technologies, including:

Hangout: A Video Conferencing Mobile App

Jun 2021 – Oct 2021

An Android video conferencing app built with Java, offering meeting codes, audio/video controls, camera switching, and chat.

Overview Hangout is a comprehensive video conferencing app designed to meet the needs of modern professionals and individuals who rely on remote communication and collaboration. The app is built using Java in Android Studio, ensuring a seamless and efficient experience for Android users. With its user-friendly interface and a wide range of features, Hangout aims to enhance virtual meetings and make them more interactive and productive.

Java