MMDT Intro to Deep Learning Projects

Applied deep-learning coursework covering tabular regression, NLP classification, CNN benchmarking, and a separately featured geospatial NLP capstone

Myanmar Data Tech Foundations of Deep Learning A+ Grade 2025

GitHub Repository · A2C Standalone Project

Overview

Completed four progressive applied AI projects across tabular regression, NLP classification, CNN benchmarking, and transformer-based geocoding. The final geospatial NLP capstone is now featured as its own project: A2C Myanmar Address-to-Coordinates.

Geospatial NLP Capstone

Built A2C Myanmar Address-to-Coordinates, a transformer-based address geocoding system trained on 610,509 Myanmar addresses with a custom Haversine distance loss and a deployed Streamlit map demo.

Earlier Projects

  • HDB Resale Price Prediction: MLP regression on Singapore open housing data; best model achieved R2=0.92 and RMSE=SGD 59,051.
  • Myanmar Location Name Classification: 24 NLP experiments across 6 architectures and 4 regularization settings; diagnosed performance limits caused by severe class imbalance.
  • CNN Benchmark on Self-Collected Dataset: Compared ResNet50, VGGNet16, InceptionV3, ConvNeXt, and EfficientNet with accuracy, inference time, model size, and parameter-count analysis.

Tech Stack: PyTorch · Hugging Face Transformers · BERT · DistilBERT · XLM-RoBERTa · Streamlit · Plotly · Kaggle GPU · Scikit-learn