Learning resources
What I read, watch, and build to stay sharp
A curated, opinionated collection of the resources that have actually moved the needle for me — organised by topic, not by hype.
How I learn
I learn best by building. Reading a paper or watching a lecture only sticks when I follow it up with a working implementation — even a toy one. Everything on this page has been through that filter.
I've tried to be honest about difficulty level and prerequisites. Some of these are genuinely hard. I'd rather give you an accurate picture than oversell something as beginner-friendly when it isn't.
Foundations of AI & ML
The bedrock. If you skip this layer, everything built on top of it feels like magic — and magic is hard to debug.
Attention Is All You Need
Vaswani et al.
The transformer paper. Dense but essential. Read it at least twice.
The Illustrated Transformer
Jay Alammar
The best visual explanation of transformers I've found. Read this before the paper.
fast.ai Practical Deep Learning
Jeremy Howard
Top-down, code-first approach. Unusually good for building intuition fast.
Neural Networks: Zero to Hero
Andrej Karpathy
Build a GPT from scratch. The best way to truly understand what's happening inside.
LLMs & prompt engineering
Working effectively with large language models — from basic prompting to structured outputs and evaluation.
Prompt Engineering Guide
DAIR.AI
Comprehensive and well-maintained. Good reference to keep open while building.
Building Systems with the ChatGPT API
DeepLearning.AI
Short, practical, and free. Good starting point for chaining LLM calls.
Anthropic's Prompt Engineering Docs
Anthropic
Unusually candid about what works and why. Applies beyond Claude.
LLM Evaluation Frameworks
Various
Hamel Husain's writing on evals is the most practical I've read on the topic.
Agents & RAG systems
Building systems that retrieve, reason, and act — the core of most of my current work.
ReAct: Synergizing Reasoning and Acting
Yao et al.
The foundational paper for ReAct-style agents. Short and readable.
LangChain Documentation
LangChain
Useful for understanding patterns even if you don't use the library directly.
Building RAG Applications
DeepLearning.AI
Covers advanced retrieval techniques — reranking, query expansion, hybrid search.
Model Context Protocol Docs
Anthropic
The spec and tutorials for MCP. Essential if you're building tool-using agents.
Production AI & MLOps
Getting AI systems out of notebooks and into production — reliability, observability, and iteration at scale.
Chip Huyen's AI Engineering
Chip Huyen
The most practical book on building production AI systems I've read in 2024–25.
Made With ML
Goku Mohandas
End-to-end MLOps curriculum. Free and genuinely comprehensive.
Weights & Biases Guides
W&B
Experiment tracking, model monitoring, and evaluation. The tooling docs are also good learning material.
The Pragmatic Engineer on AI
Gergely Orosz
Grounded, sceptical takes on AI in engineering orgs. Good antidote to hype.
Books I'm Reading & Recommend
The books currently on my desk or recently finished — each one worth the time if you're serious about building with AI.
Domain-Specific Small Language Models
Guglielmo Iozzia
Practical guide to fine-tuning, quantizing, and deploying SLMs on commodity hardware. Cuts through the hype — if you want capable AI without cloud GPU bills, start here.
Agentic Architectural Patterns for Building Multi-Agent Systems
Dr. Ali Arsanjani & Juan Pablo Bustos
Enterprise-grade design patterns for agentic AI — covers RAG, LLMOps, A2A protocol, and how to move from prototype to production. Written by the Director of Applied AI Engineering at Google Cloud.
Multi-Agent Systems Engineering
Cannon T. Hale
The only book in the category that gives you quantified thresholds for architecture decisions. Covers centralized, peer, and market coordination models with benchmarks, fault injection, and a production readiness checklist.
30 Agents Every AI Engineer Must Build
Imran Ahmad
Hands-on pattern library for production agent architectures using LangChain and LangGraph. Covers memory, planning, reasoning, tool use, and multi-agent orchestration with real code.
AI in Human Terms
David Lloyd
The best book I've found for explaining AI to non-technical people — no math, no jargon. Useful for anyone who needs to communicate what they're building to stakeholders who aren't engineers.
Learning by building
The best way to learn is to ship something. The Projects section shows what I've built while working through these resources.