Data Scientist

Technical Skills: Python, SQL, AWS, Microsoft Power BI, R, MongoDB

Education

• M.S (Data Science) | Universiti Teknologi Malaysia (Mar 2023-Jul 2024)
• B.S Petroleum Chemistry with Honours| Universiti Putra Malaysia (Sep 2018-Jul 2022)

Projects

Student Performance Prediction

GitHub

Developed a machine learning model for predicting student performance in Mathematic, leveraging various factors to provide early intervention strategies using Python. Implemented a Flask web application to deliver insights through a user-friendly interface, and deployed it on AWS Cloud for scalability and reliability. This project led to more accurate identification of at-risk students, demonstrating the potential of data-driven approaches in educational outcomes. The automated process enhances decision-making by offering precise predictions and actionable insights.

Chicken Disease Classification

GitHub

Developed a deep learning model for Chicken Disease Classification using TensorFlow and Keras. The project involved creating an end-to-end pipeline that ingested data from GitHub, trained a VGG16 model on images of chicken feces, and deployed the model on AWS using GitHub Actions for CI/CD. The model achieved high accuracy in distinguishing between healthy and infected samples, demonstrating the effectiveness of transfer learning in medical diagnostics. This work highlights the potential of leveraging image classification to enhance poultry health management and streamline disease detection processes.

Cooking Oil Price Prediction in Malaysia

GitHub Project Overview: Developed and validated time series forecasting models to predict grocery prices using a variety of techniques such as SARIMA and LSTM neural networks. The project focused on identifying key trends and seasonal patterns in daily price fluctuations, providing accurate short-term and long-term price forecasts.

Achievements:

• Achieved a 15% improvement in Mean Absolute Percentage Error (MAPE) by optimizing SARIMA and LSTM model parameters.

• Built a multi-step data pipeline to handle over 100,000 records, automating data cleaning and feature engineering, resulting in a 30% reduction in preprocessing time.

• Implemented a scalable LSTM model with Keras, utilizing early stopping and hyperparameter tuning, leading to a 25% reduction in training time and 8% higher forecast accuracy.