Resources

Fastest path to useful Python for non-CS people — you are writing real scripts within hours, not learning abstract theory.

The authoritative reference — dry, but covers the language precisely; use it to look things up, not to learn from scratch.

The canonical Git reference — chapters 1–5 are all most practitioners ever need; the rest covers internals for the curious.

Fills the gap every formal program leaves: the shell, Vim, tmux, Git, debugging, profiling — tools you use every day but were never taught.

Solid foundational drilling — the exercises are genuinely useful, though the content stops well short of inference at the level you need for professional work.

A genuinely good free statistics textbook — readable, covers inference properly, and has real exercises; the free PDF is the real deal, not a sample.

The best bridge between statistics and data science practice — it treats you as a programmer who needs to think statistically, not a stats student who needs to code.

The fastest way to build and publish interactive dashboards; the free tier is genuinely capable and the gallery shows you what is possible.

The rigorous big brother of ISLR — dense, mathematically demanding, and necessary if you want to understand WHY the algorithms work at a deep level.

The best practical ML book — clear code, real projects, and it takes you all the way from scikit-learn to neural networks in one coherent progression.

The single best bridge between undergraduate mathematics and ML — covers linear algebra, calculus, and probability in the context of ML applications.

The canonical first ML course — Ng explains concepts with unusual clarity, though the assignments are now Python-based and more hands-on than the original.

The theory canon of deep learning — rigorous, dense, and essential for understanding WHY neural networks work; not a first read, but an important eventual reference.

You build a transformer from scratch — every gradient, every matrix multiply — and you understand it by the end; this is the closest thing to an essential ML video series.

The best top-down practical DL course — starts with working code and walks backward to the theory; covers vision, NLP, and tabular data.

The production MLOps bible — covers every step from data collection to monitoring in production, with clear frameworks and real-world cases; non-negotiable for this stage.

The official, authoritative PyTorch reference — the introductory tutorials are the fastest way to get productive; use the advanced sections for specific capabilities.

Bridges the gap between training a model and shipping a product — covers tooling, infrastructure, and the product decisions around ML.

The official Docker introduction — clear, practical, and the right starting point for containerising ML workloads.

Bite-sized and current — the RAG, agents, and evals courses are the fastest way to get hands-on with specific LLM engineering topics.

A minimal, readable GPT implementation from scratch — the best codebase to read when you want to understand what is actually happening inside an LLM.

Exceptional deep-dives on agents, LLMs, and hallucination — among the best technical writing in ML; dense but worth every word.

The authoritative reference for the Claude API, prompting best practices, and agent architecture — if you are building with Claude, start here.

The standard RAG evaluation framework — covers faithfulness, answer relevancy, and context precision in a way you can actually integrate into a CI pipeline.

The primary framework for multi-agent orchestration with explicit state — useful for building agents that need complex branching or human-in-the-loop.

Clear, practical docs for one of the best vector databases — the quickstart gets you a working similarity search in minutes.

The paper that established masked language modelling as the pretraining paradigm — still the canonical read for understanding BERT-family models.

The GPT-3 paper that demonstrated in-context learning at scale — important for understanding why LLMs behave the way they do.

The paper that formalised chain-of-thought prompting — short, clear, and essential for anyone doing reasoning-heavy LLM work.

The original RAG paper — explains the retrieval-generation architecture that underpins most production LLM knowledge systems today.

The paper that made fine-tuning accessible — introduces the low-rank decomposition trick that is now the standard PEFT method.

The InstructGPT / RLHF paper — explains how models are aligned with human preferences; essential for understanding why modern LLMs behave so differently from raw language models.

The Bayesian-flavoured ML canon — rigorous, comprehensive, and now free from the author; the probabilistic graphical model chapters are the best treatment available.

The modern research reference — encyclopaedic, up-to-date through 2022, and freely available; Vol. 1 covers foundations, Vol. 2 covers advanced methods.

The best place to find state-of-the-art results with reproducible code — benchmark tables, leaderboards, and linked implementations in one place.

An archive of exemplary ML explanation — interactive, visually precise, and a high bar for what clear technical communication looks like; no longer actively publishing, but everything there is worth reading.

The preprint firehose — every important ML paper appears here before any journal; the skill is learning to filter signal from noise.

Builds the intuition for derivatives and integrals that most textbooks never give — the chain rule visualisation alone is worth the series.

The canonical linear algebra course — Strang's lectures are legendary and the exercises are genuinely instructive; this is the course that makes linear algebra click.

The optimisation reference — comprehensive, mathematically rigorous, and free; most practitioners need chapters 1–5; researchers need the rest.

The best probability course and textbook combination — Blitzstein's lectures are the clearest treatment of conditioning, distributions, and Bayes available for free.