Math foundations
Only the maths that shows up in interviews and in debugging. Operational understanding, not proofs.
| # | File | The question it answers |
|---|---|---|
| 01 | Linear algebra for ML | why attention is quadratic, why embeddings get normalised, what LoRA factorises |
| 02 | Probability and statistics for ML | why a 99%-accurate fraud model is useless, what a p-value isn’t, why p99 not mean |
| 03 | Calculus and optimisation | why your loss went NaN, Adam vs AdamW, why fine-tuning needs 4x the memory |
| 04 | Information theory for ML | where cross-entropy comes from, what perplexity means, what temperature does |
If you only read one, read 02 — the base-rate calculation is the single most reusable piece of statistics in an ML interview.