From regression to retrieval-augmented generation

For engineers who want to actually understand what is happening inside the model — not just call an API. Linear regression by hand on a whiteboard all the way to fine-tuning a small open-source LLM and serving it behind a retrieval pipeline.

Curriculum

1. Foundations. Linear algebra refresh, gradient descent, and bias / variance trade-offs you can feel.
2. Classical ML. Regression, trees, gradient boosting, and clustering using scikit-learn.
3. Deep learning. PyTorch, autograd, training loops you write yourself.
4. Sequence models. RNNs, attention, and transformers — implemented from scratch in 200 lines.
5. Modern LLMs. Prompt engineering, evaluation, fine-tuning with LoRA, RAG with pgvector.