MY Projects
Image Classification using MobileNetV2
This project focuses on building and evaluating a deep learning–based image classification system to categorize natural scene images into distinct classes such as buildings, forests, and streets. Using convolutional neural networks, I implemented data preprocessing, model training, and performance evaluation pipelines to assess classification accuracy and generalization. Beyond model development, the project emphasizes translating model outputs into structured insights that could support automated inspection, environmental analysis, or decision-support systems where large volumes of visual data must be analyzed efficiently and consistently.
Sensor Tower Web Scraper | Python, Playwright, AsyncIO
Designed and implemented an automated web scraper using Playwright to extract publisher country information for over 8,000 mobile apps from Sensor Tower. The script efficiently navigates dynamic content behind a login wall, mimics human-like browsing behavior with randomized delays and periodic breaks to avoid rate limiting, and includes robust error handling and resume functionality. Leveraging asynchronous programming with asyncio, the tool achieves scalable and reliable data extraction while ensuring session stability and progress persistence. This project demonstrates proficiency in browser automation, real-world data collection at scale, and ethical scraping practices.
Rubiverse – Your Guide to Rubina Almas 🌌
Rubiverse is my digital twin — an intelligent assistant that introduces who I am, what I’ve built, and the value I bring to organizations. Powered by 'gpt 4.1 mini' and a custom context engine, Rubiverse transforms a traditional résumé into a conversational, interactive experience. It engages recruiters like a friendly career storyteller while automating résumé delivery and offering a smooth, modern UI. This project reflects my passion for blending AI, design, and meaningful communication to create innovative user experiences.
Retail Stockout Risk Prediction
This project uses machine learning to help retailers avoid running out of popular products. By analyzing demand forecasts, sales behavior, inventory levels, pricing, and seasonal trends, the model identifies which items are at high risk of stockout.
To handle data imbalance, SMOTE oversampling was applied, and an XGBoost classifier was trained, achieving strong prediction performance.
This proactive approach helps retailers reduce lost sales, improve product availability, and make smarter inventory decisions , ultimately creating a better shopping experience for customers.
SafeSphere- A Multimodal AI for Identifying and Alleviating Hate Speech
This project created a Python-based Instagram scraper to gather user interactions, reel metadata, and comments. Using machine learning methods, the data was analyzed for sentiment analysis and multilingual hate speech identification. NLP, web scraping, and sophisticated visualization techniques are all included into the project.
Analysis of Arizona Businesses
This project analyzes Yelp data to uncover trends in local business performance and user behavior across Arizona. Using distributed computing frameworks such as Hadoop and Apache Spark (via PySpark), the project involved merging multiple large-scale JSON datasets, transforming them into a queryable format (Parquet), and running Spark SQL queries to extract insights. Two milestones were completed: the first focused on business-level analysis (e.g., identifying high-rated categories and locations), and the second on user-level behavior (e.g., activity levels, review sentiment, and community influence). Results were visualized and reported in Jupyter notebooks.
Forecasting 2025 Job Market Using ARIMA
This project uses an ARIMA model to forecast the unemployment rate in the United States for the year 2025. The model is built using historical unemployment data and interest rates, focusing on the period from 2000 to 2022. The project involves data cleaning, visualization, and time series forecasting.