Prerequisites
- Python fluency (this is the floor).
- Basic linear algebra and calculus (Form 5 level is enough to start).
- Probability basics.
Phase 1 — Classical ML
- scikit-learn: regression, classification, clustering.
- Understand overfitting, cross-validation, metrics.
- Project: predict HK weather or DSE grades from features.
Phase 2 — Neural Networks
- PyTorch basics: tensors, autograd, optimisers.
- Simple MLP on MNIST.
- Loss functions and regularisation.
Phase 3 — Deep Learning
- CNNs for images.
- RNNs/Transformers for text.
- Transfer learning with Hugging Face.
Phase 4 — Applied LLMs
- Prompt engineering with OpenAI and Anthropic APIs.
- Retrieval-augmented generation.
- Fine-tuning small open models.
Where to Learn
- fast.ai — the gold-standard course, free.
- Andrew Ng's Coursera specialisation.
- HK AI society meetups at HKUST and HKU.
Practise this on PyForm — free
PyForm runs Python in your browser with an AI tutor trained for HKDSE. No install, no credit card.
Open PyForm →