These are the books and papers that any serious data scientist and AI practitioner should know. Not every paper ever published — the ones that actually changed the field and will change how you think about it.
Books — Read These First
The Elements of Statistical Learning — Hastie, Tibshirani, Friedman
The mathematical foundation of supervised and unsupervised learning. Dense, rigorous, essential. Available free from the authors. Use as reference, not linear read.
An Introduction to Statistical Learning — James, Witten, Hastie, Tibshirani
The accessible companion to ESL. Start here if ESL’s mathematics is currently out of reach. Also free from the authors.
Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow — Aurélien Géron
The best practical ML book for Python practitioners. Covers the full stack from classical ML through deep learning. The book to read if you want one book that bridges theory and practice.
Data Science for Business — Provost & Fawcett
Rare in that it explains the business context for analytical methods alongside the methods. Essential for practitioners who need to communicate the value of analytical work to non-technical stakeholders.
Designing Machine Learning Systems — Chip Huyen
The production ML book. Covers data pipelines, feature stores, deployment, monitoring, and the organisational realities of keeping ML systems alive in production. Nothing else covers this as comprehensively.
Foundational Research Papers
Attention Is All You Need (Vaswani et al., 2017) — The transformer paper. Everything in modern NLP and a significant portion of modern AI traces back here. Read it.
BERT: Pre-training of Deep Bidirectional Transformers (Devlin et al., 2018) — How bidirectional pre-training changed NLP. The architecture that preceded GPT-style models in industrial adoption.
Language Models are Few-Shot Learners (Brown et al., 2020) — The GPT-3 paper. Demonstrates emergent few-shot learning at scale. The paper that changed how the industry thought about what large models could do without fine-tuning.
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., 2020) — The RAG paper. Foundational for anyone building LLM applications that require grounding in external knowledge.
On AI in the Real World
The Alignment Problem — Brian Christian
The clearest non-technical account of what it actually means to align AI systems with human values. Well-researched and genuinely useful for anyone thinking about AI governance or responsible deployment.
Atlas of AI — Kate Crawford
On the material and political dimensions of AI systems — the data, the labour, the infrastructure, and the power. Important counterweight to purely technical framings.