Projects

Climate Event Predictor

A machine learning project that analyzes global disaster data to predict both climate event severity and economic impact. This project demonstrates a complete ML pipeline — including EDA, feature engineering, outlier treatment, classification, regression modeling, and performance interpretation.

The dataset includes worldwide climate events with attributes such as event_type, country, duration, fatalities, injuries, infrastructure_damage, economic_loss, and more. Over 15 engineered features including casualty_rate, damage_per_day, severity×duration, and population_density_proxy significantly strengthened model performance.

Key Insights:
- Severity prediction is highly accurate (Random Forest: 90.33%)
- Economic loss data contained extreme outliers (up to $700M)
- Capping outliers at the 99.5th percentile improved R² from 0.0864 → 0.7960
- Final Gradient Boosting model explains ~80% of economic variance

Final Model Results:
- Severity Classification Accuracy: 90.33%
- Economic Impact Regression R²: 0.796

Skills: Python, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, Random Forest, Gradient Boosting, Outlier Treatment, Classification & Regression Modeling

Github

House Price Predictor

A machine learning project that predicts house prices using Linear Regression with features such as square footage, number of rooms, location, and house quality. This project showcases an end-to-end ML workflow—covering data cleaning, feature engineering, modeling, evaluation, and interpretation.

The model was trained on a housing dataset containing attributes like sqft_living, sqft_basement, bedrooms, bathrooms, grade, view, zipcode, and more. Engineering new features such as price_per_sqft, house_age, total_sqft, and total_bathrooms significantly improved performance.

Final Model Results:

- MAE = $52,251
- RMSE = $80,107
- R² = 0.91

Cross-validation showed stable performance with an average R² of 0.91. Visualizations like correlation heatmaps, scatterplots, boxplots, and residual plots were used for deeper insights.

Skills: Python, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, Feature Engineering, Linear Regression, Cross-Validation

Github

Email Spam Classifier

A natural language processing (NLP) project that classifies emails as Spam or Not Spam using TF-IDF vectorization and Logistic Regression. The model is deployed on Streamlit, enabling real-time prediction based on user-entered email content.

The dataset used for training was a cleaned and merged version of the Complete SpamAssassin Corpus, containing thousands of real-world emails. The model leverages text preprocessing, n-gram feature extraction, and a pipeline-based architecture to ensure efficiency and reproducibility.

Model Pipeline:

- TF-IDF Vectorizer (1–2 grams, 8,000 features)
- Logistic Regression (max_iter = 1000)

Model Performance:

- Accuracy = 95.3%
- Precision = 0.899
- Recall = 0.958
- F1 Score = 0.927

Confusion matrix showed strong recall, indicating that the model catches most spam emails while keeping false negatives low.

Skills: Python, Pandas, Scikit-learn, NLP, TF-IDF, Logistic Regression, Confusion Matrix Analysis, Streamlit Deployment

Github Live App

Academic Performance Regression Model

A Machine Learning project focused on predicting students’ final exam scores using the UCI Student Performance Dataset. This regression model estimates a student's final grade (G3) by analyzing factors such as study habits, alcohol consumption, support systems, absences, and their previous academic performance.

The workflow involved:

• Data Preparation: Cleaning the dataset, engineering features (like avg_previous_grade and weekend alcohol consumption), handling categorical variables through one-hot encoding, and scaling numerical features for consistent model performance.

• Modeling: Implemented and compared multiple regression models including:
- Linear Regression (baseline model)
- Polynomial Regression (degree 2)

• Evaluation: Assessed performance using MAE, RMSE, and R² scores. Visual insights included correlation heatmaps, predicted vs. actual plots, and residual diagnostics to evaluate model fit.

Results:
- Baseline R²: 0.752
- Polynomial R²: 0.757
- Average prediction error: ~1.6 grade points

The strongest predictor was a student’s prior grades, while behavioral factors (study time, absences, alcohol use) showed weaker but meaningful influence.

This project demonstrates practical educational analytics, showing how ML can help identify patterns in academic performance and support student-success strategies.

Skills: Python, Pandas, NumPy, Scikit-learn, Regression Modeling, Data Visualization, Feature Engineering

Github

Pymol Copilot

Created by Arqam Usman Ali, Abdul Lateef FNU, Saipranith Oku, and Sriram Koyalkar

-- NJIT YWCC Spring 2024 Capstone 3rd Place Panel Winner --

For my senior capstone project at the New Jersey Institute of Technology (NJIT), I led a team of four to develop an AI Copilot for PyMOL, a drug visualization software widely used in biology labs. The project was inspired by my roommate, who worked in NJIT’s biology lab and faced challenges mastering PyMOL’s complex interface, which typically required 4–6 weeks of training for new hires.

To address this issue, we envisioned a solution leveraging GPT-4's API to simplify workflows and reduce the learning curve. By enabling users to interact with PyMOL through natural language commands, the AI Copilot translated user input into backend functions, making the software more accessible and significantly enhancing its usability. The final product streamlined molecule visualization, accelerating viewing processes by 6 times and property analysis by 2 times, as reported by researchers.

In my role as project manager and backend developer, I guided the team through four months of development, organizing sprints, conducting daily scrum meetings, and ensuring timely progress. I also played a key role in implementing the backend by training GPT-4’s API in Python to map natural language inputs to PyMOL's plugin library functions. While a teammate focused on creating a user-friendly chat interface with PyQt, I collaborated closely with the backend team to ensure seamless integration and functionality. Our project’s success was highlighted by earning third place at the Capstone Showcase, underscoring the innovation and impact of our work.

Skills: Python, GPT-4 API, PyMOL, PyQt

Github

Tic-Tac-Toe

Created by Arqam Usman Ali

A browser-based Tic Tac Toe game designed for interactive play between two users. This project implements the classic game using HTML, CSS, and JavaScript, allowing players to take turns clicking on cells to place Xs and Os. The game dynamically detects wins, draws, and provides a responsive interface for smooth gameplay.

The application leverages DOM manipulation to update the game board in real-time, ensuring an engaging user experience. With built-in logic to check for winning conditions and game resets, players can enjoy multiple rounds without refreshing the page.

Skills: HTML, CSS, JavaScript, DOM Manipulation, Problem Solving, Debugging

Github Live App

Book Recommendation Engine

Created by Arqam Usman Ali

A web-based Book Recommendation Engine designed to help users discover their next favorite book effortlessly. This application suggests books based on genre, allowing users to explore top-rated titles or the latest releases.

Users can enter a book title, and the engine fetches recommendations, sorted by rating or publication year for a personalized reading experience. The intuitive interface ensures seamless navigation and a visually appealing layout.

Skills: JavaScript, HTML, CSS, JSON Handling, Data Processing

Github Live App

Harry Potter Quiz

Created by Arqam Usman Ali

An interactive web-based quiz designed for Harry Potter fans to test their wizarding knowledge. This project presents a dynamic quiz with multiple-choice questions, providing instant feedback based on the user's score.

Players can answer 10 magical questions and receive personalized results, ranging from Muggle status to potential Hogwarts valedictorian. The quiz features an enchanting Harry Potter-themed UI, bringing the magic of the wizarding world to life.

Skills: JavaScript, DOM Manipulation, Event Handling, JSON Data Fetching, HTML, CSS

Github Live App

Choose Your Own Adventure Game

Created by Arqam Usman Ali

An interactive text-based adventure game where players navigate through a branching storyline by making choices that lead to different outcomes, including victory, survival, or loss. This project dynamically updates text, images, and buttons based on player decisions, creating an engaging and immersive experience.

The game features multiple story paths, allowing players to explore different narratives based on their choices.

Built with JavaScript, HTML, and CSS, this project demonstrates the power of dynamic storytelling and user interaction without the need for complex frameworks.

Skills: JavaScript, HTML, CSS, Game Development, Interactive UI

Github Live App

Discord Message Automation

Created by Arqam Usman Ali

A software development project that automates the process of sending random messages to a Discord channel. The project integrates the Discord API, utilizes Python libraries like `Faker` to generate realistic random content, and implements automated testing with `pytest`. The solution includes a structured directory setup with folders for the main application logic (`discordmsg`) and automated testing (`tests`). Key features include dynamically generated payloads, API integration, and robust testing to validate functionality.

Whether you're looking to automate tasks in Discord or learn how to work with APIs and Python testing frameworks, this project offers a practical, hands-on experience. Dive into the code to explore a real-world example of API usage, random data generation, and test-driven development!

Skills: Python, Discord API, Faker, Pytest, Dotenv, Requests

Github

Task Trackr

Created by Arqam Usman Ali

A mobile-responsive web application designed to help users manage and track tasks with ease. This project implements a task tracker using CRUD operations, allowing users to add, complete, and delete tasks. Built with Next.js, it features an intuitive interface that updates dynamically without requiring page reloads.

Users can interact with the task list by clicking on tasks to mark them as completed and remove tasks they no longer need. The mobile-responsive design ensures a seamless experience across devices.

Ideal for anyone looking to learn about building interactive web applications with Next.js, this task tracker application offers a hands-on experience in task management, React, and web design.

Skills: Next.js, React, JavaScript, Tailwind CSS

Github Live App

Calculator Website

Created by Arqam Usman Ali

A fully functional calculator website developed using HTML and CSS, designed to perform all basic arithmetic operations such as addition, subtraction, multiplication, and division. The project incorporates flash message validation to ensure user-friendly error handling, providing alerts for empty input fields and checks for division by zero.

Built and tested on PyCharm, the application ensures reliability and correctness by utilizing Pylint for static code analysis. Additionally, continuous integration testing was implemented using Travis CI to maintain code quality and streamline the development process.

This project highlights the application of fundamental web development techniques combined with robust testing methodologies, ensuring a seamless user experience and reliable performance.

Skills: HTML, CSS, PyCharm, Pylint, Travis CI, Frontend Validation

Github

Video Demonstration

Weight Converter

Created by Arqam Usman Ali

A responsive weight converter application designed using JavaScript and Bootstrap. This project allows users to convert weight from pounds to grams, kilograms, and ounces instantly. With its intuitive interface and real-time conversion functionality, the application demonstrates practical use of JavaScript event handling and Bootstrap's styling capabilities.

This project is ideal for showcasing proficiency in building practical web tools that integrate modern design principles and client-side scripting.

Skills: JavaScript, Bootstrap, DOM Manipulation, Responsive Design

Github

Video Demonstration

Lexicon-Based Sentiment Analyzer

Created by Arqam Usman Ali

This project involved developing a sentiment analysis tool in R using the AFINN lexicon, designed to analyze textual data with precise negation handling for enhanced accuracy. The tool processed large datasets to measure sentiment proportions, such as positive, negative, and neutral expressions, while providing insights into emotional trends, user feedback, and overall sentiment tone. By visualizing sentiment distributions, the analyzer revealed patterns in textual data, enabling strategic decision-making and targeted improvements to enhance user engagement and understanding.

Skills: R, Sentiment Analysis, Data Visualization

Github

JavaScript Slideshow

Created by Arqam Usman Ali

A dynamic JavaScript slideshow designed to display a series of images seamlessly. This project leverages core JavaScript functionality to create a smooth and automated image transition effect. The slideshow cycles through an array of images at a predefined interval, offering a visually engaging experience for users.

The slideshow features a lightweight implementation with a responsive image display. It is initiated with a simple `onload` event and uses JavaScript's `setTimeout` function to transition images. This project demonstrates foundational JavaScript skills and showcases the ability to implement interactive and engaging web components.

Perfect for learning how to manipulate the DOM and work with JavaScript arrays, this project is an excellent example of creating interactive web elements without relying on external libraries.

Skills: JavaScript, HTML, DOM Manipulation, Responsive Design

Github Video Demonstration

Product Card UI Design

Created by Arqam Usman Ali

A sleek and modern product card designed to showcase UI design skills using HTML and CSS. This project features a responsive product card with interactive elements such as size and color selectors, product ratings, and pricing details. The design highlights attention to detail with clean typography, engaging visuals, and smooth layout organization.

This project is perfect for demonstrating proficiency in front-end development and an eye for design, offering a practical example of how to build engaging user interfaces.

Skills: HTML, CSS, Font Awesome, Responsive Design

Github

Video Demonstration

Activities

NJIT Students for Hunger Relief

Students for Hunger Relief (SFHR) is a student-run organization at NJIT dedicated to addressing hunger in Newark through large-scale, on-campus initiatives. SFHR collaborates with charitable organizations to acquire food and organizes meal packaging events where hundreds of meals are prepared by volunteers. These meals are then distributed to shelters and locations in Newark such as Penn Station, Lincoln Park, and Covenant House, helping to connect NJIT students with the local community.

I was an Volunteer Coordinator for Students for Hunger Relief from 2022 to 2024. During this time, I helped organize and manage 15 meal packaging events, resulting in over 1,800 meals being distributed to homeless shelters in Newark.

My responsibilities included coordinating logistics, managing volunteers, and ensuring smooth operations during the events. The meals packaged ranged from rice and chicken to beef chili and salads, all sourced from 501(c)(3) accredited organizations.