🤖 An Artificial Intelligence/ Deep Learning / Machine Learning Enthusiast

👨‍🎓 Bachelor of IT in Data Science (2022 - 2025)

🏫 Master of IT in Artificial Intelligence (2025 - 2026)

• Dataset: CIFAR-10 dataset used for generative image synthesis consisting of 60,000 color images across 10 categories (airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck).

• Model Implemented:

  • Denoising Diffusion Implicit Model (DDIM) for image generation.
  • Custom U-Net architecture with attention mechanisms.
  • Implemented both forward (noise addition) and reverse (denoising) diffusion processes.

• Achievement:

  • Demonstrated stable convergence and interpretable attention patterns over 100 training epochs.
  • Evaluated generated samples with Fréchet Inception Distance (FID = 102.02) and Inception Score (IS = 1.21), showing trade-offs between sampling speed, image fidelity, and diversity.

• Future Improvement:

  • Implement conditional diffusion guidance for class-controlled generation.
  • Refine noise scheduling for improved efficiency and sample quality.
  • Experiment with deeper U-Net or transformer-based architectures to enhance robustness and image realism.

• Dataset: A private car sales dataset from a Kaggle competition containing vehicle attributes such as year, mileage, brand, fuel type, and engine specifications to predict selling prices.

• Model Implemented:

  • Applied supervised regression models including RandomForest, XGBoost, LightGBM, and Gradient Boosting.
  • Focused on data preprocessing, feature engineering, and hyperparameter optimization using GridSearchCV.
  • Utilized Recursive Feature Elimination (RFE) to identify the most significant predictors of price.

• Achievement:

  • Developed a car price prediction model achieving 10% Mean Absolute Percentage Error (MAPE).
  • Compare the capabilities and the peformance of each model to provide the highest accuracy.

• Applied K-Means++ and Agglomerative Clustering to segment customers based on purchasing behavior.
• Enabled businesses to design personalized marketing strategies, improve customer retention, and optimize product recommendations.
• Conducted feature scaling and elbow/silhouette analysis to determine optimal cluster numbers.
• Compared the clusters generated from two algorithms (see how they perform differently, we don't measure accuracy here 🙂).
• Recommend strategies to target each customer segment generated from the two algorithms.