Portfolio
Portfolio
Previous Project
Here are selected projects that showcase my expertise in machine learning, AI, and data-driven problem solving
Computer Vision Object Recognition (Kitchen Dataset)
Trained a YOLOv8 model on a custom kitchen objects dataset.
Annotated over 7,000 images and achieved strong precision on player detection, though performance was lower on small objects (like the ball).
Demonstrated model training, evaluation, and visualization of detection results.
Tools: Python, YOLOv8, Roboflow, Google Colab
Credit Risk Prediction System
Developed a machine learning model that predicts whether a loan applicant is likely to default or repay based on demographic, financial, and behavioral data.
Built with Python, Scikit-learn, and FastAPI, the model uses Logistic Regression and Random Forest algorithms to evaluate applicant creditworthiness, achieving 89% accuracy and a ROC-AUC of 0.85.
Deployed as a live interactive web app with a connected frontend, it allows users to input data and receive real-time credit risk predictions, supporting data-driven loan decisions.
Chatbot Development using Prompt Engineering
Designed a football-themed chatbot using GPT-4 prompt engineering. Focused on dialogue flow, tone consistency, and contextual understanding to create an engaging, human-like conversation experience.
Tools: Python, OpenAI GPT-4 API, Prompt Engineering
Student Grade Prediction
Created a machine learning model to predict student grades using Kaggle datasets. Applied SVM algorithms, data normalisation, feature scaling, and cross-validation to ensure reliable predictions.
Tools: Python, Scikit-learn, Kaggle Datasets
Hotel Booking Forecast
Developed a predictive model using time series and machine learning techniques to forecast booking demand, helping optimize resource planning.
House Price Prediction
House Price Prediction Model — Built a regression-based ML model using Python (Scikit-learn, Pandas, NumPy) to predict house prices with high accuracy by leveraging key features such as location, size, and number of rooms.
Sports Performance Analysis with Computer Vision
Built an open-source pipeline to track football players from raw match footage using YOLOv8 and DeepSORT.
Extracted metrics such as speed, distance, workload, and intensity, and used them to analyze tactical patterns like counterattacks and defensive transitions.
Showed how video-based analysis can offer affordable, scalable insights for grassroots and semi-pro teams without GPS trackers.
Tools: Python, YOLOv8, DeepSORT, OpenCV, Pandas, Matplotlib, Roboflow
Bank Customer Churn Prediction
Developed a machine learning system that predicts whether a bank customer will leave or stay based on their financial and behavioral data.
Built with Python, Scikit-learn, and FastAPI, the model was trained using a Random Forest Classifier, achieving 87% accuracy and a ROC-AUC of 0.86.
After optimizing the decision threshold to 0.57, performance improved to 90% accuracy, balancing precision and recall.
Deployed as a live FastAPI web app integrated with a frontend interface, it provides real-time churn predictions, with Age identified as the strongest factor influencing customer churn.
Hate Speech Detection Using Machine Learning
Built a text classification model to detect hate speech using NLTK and a Decision Tree Classifier. Applied tokenisation, stop-word removal, and feature engineering to improve model accuracy.
Tools: Python, NLTK, Scikit-learn
Spam Mail Prediction
Implemented a spam detection system using Logistic Regression. Pre-processed and vectorised email text data, and evaluated model performance with precision, recall, and F1-score.
Tools: Python, Scikit-learn, Pandas
Statistical Test on Productivity and Time Management
Conducted a correlation and hypothesis testing analysis to study the link between productivity and time management. Delivered data-driven insights on how time allocation impacts performance.
Tools: Python, Excel, Statistical Testing
Content Optimization with TikTok Data
Applied mathematical optimisation techniques to design a posting strategy that maximized reach and engagement for AI-focused TikTok content
Customer Sentiment Analysis
using Python to classify and visualize customer reviews as positive, negative, or neutral, providing insights into satisfaction trends.