About Me

Hello, I am Joy, a passionate Applied AI Solutions Development student at George Brown College, driven by the power of data to solve real-world challenges. My journey combines hands-on technical expertise with a commitment to transforming complex datasets into actionable insights that create positive business effect. With proficiency in Python, SQL, PySpark, and Microsoft Azure, I develop end-to-end AI solutions that encompass ETL pipelines, predictive models, and intelligent chatbots. My academic background in Agile methodologies, Big Data Tools, and Machine Learning enables me to excel in fast-paced, collaborative environments.

In my Data Management and Reporting project, I designed an ETL workflow to clean and load Superstore sales data into a normalized MySQL database, producing executive dashboards that guided strategic decisions. I also built an Enterprise Knowledge Assistant a RAG-powered chatbot for NASA policy documents using NLP, transformer embeddings, and vector search to deliver accurate, context-aware responses. These projects highlight my ability to manage the full data lifecycle, from ingestion and preprocessing to modeling, visualization, and deployment.

My Final Project in Big Data Techniques further demonstrates this through the analysis of large-scale public safety data using SSAS, Alteryx, Spark, and Power BI to uncover trends and support decision-making. This portfolio showcases selected projects that reflect both my technical expertise and my problem-solving mindset, offering deeper insight into how I approach challenges with creativity, precision, and purpose.

KNOWLEDGE ENTERPRISE MANAGEMENT WITH RAG

The Enterprise Knowledge Assistant is a Retrieval-Augmented Generation (RAG) chatbot designed to provide intelligent, context-aware access to NASA policy documents. It integrates LangChain, SentenceTransformer, and ChromaDB to combine semantic retrieval. The system ingests and processes PDF documents, transforms them into embeddings, and stores them in a vector database for fast, relevant retrieval. Queries are converted into semantic vectors and matched against the stored corpus to extract the most relevant information. The retrieved context is then passed to a language model to generate accurate, grounded responses. Built using python, it demonstrates a full-lifecycle AI development from data ingestion to intelligent response generation and an expertise in NLP, vector search, and LLM integration. It enhances enterprise knowledge management by turning unstructured text into actionable insights. This project exemplifies scalable, explainable AI in document intelligence and retrieval systems.


SEE MORE AT: https://github.com/musajoy4/RAG-pipeline


Click Here

BBCNEWS TOPIC CLASSIFICATION WITH NLP

A machine learning project that classifies BBC news articles into five categories tech, business, sport, entertainment, and politics using NLP techniques.

It loads and explores a 2,225-article dataset, revealing balanced categories and text characteristics.

Text preprocessing includes cleaning, lemmatization, and TF-IDF vectorization with optimized feature selection. Multiple models are trained and compared, with hyperparameter-tuned Logistic Regression achieving 97.18% accuracy.

See full details here: https://github.com/musajoy4/topic-classification

Click Here

CUSTOMER CHURN PREDICTION

This project focuses on building a churn prediction system using the IBM Telco dataset to help telecoms retain revenue. It engineered features from demographics, services, and billing data while training and comparing various models such as Logistic Regression, Random Forest, XGBoost, and SVM models. The best model achieved 81% recall and 84% ROC-AUC, catching 106 additional churners and unlocking $53,000 in potential savings.

Explore the full project: https://github.com/musajoy4/customer_churn_prediction
Click Here

About

Lorem ipsum dolor sit amet, consectetur et adipiscing elit. Praesent eleifend dignissim arcu, at eleifend sapien imperdiet ac. Aliquam erat volutpat. Praesent urna nisi, fringila lorem et vehicula lacinia quam. Integer sollicitudin mauris nec lorem luctus ultrices. Aliquam libero et malesuada fames ac ante ipsum primis in faucibus. Cras viverra ligula sit amet ex mollis mattis lorem ipsum dolor sit amet.