Phase 3 · 2–3 weeks
NLP Fundamentals
From word vectors to attention — and why attention won.
What this phase teaches
The journey from word-level to subword tokenization (BPE, WordPiece, Unigram, SentencePiece), word/contextual embeddings, n-gram and RNN/LSTM/GRU language modeling, seq2seq with attention (Bahdanau, Luong).
Anchor resources
- Stanford CS224N (Manning)
- Hugging Face LLM Course chapters 1–4
- Speech and Language Processing (Jurafsky & Martin — free online)
- Papers: word2vec (1301.3781), seq2seq (1409.3215), Bahdanau attention (1409.0473)