AI & Machine Learning resources for developers

A nine-stage path for choosing the model, hosting and tools for an AI feature on evidence rather than hype: requirements first, reasoning versus fast models, cost per task, context and retrieval, tool calling and MCP, API versus open-weight models, the rest of the stack, a three-model comparison on your own data, and routing and fallbacks in production. It suits developers, students and founders about to build with an LLM. By the end you can defend a model choice with your own numbers and switch in a day.

• AI
• Backend

A nine-stage path for securing LLM apps and AI agents, or testing them as a red-teamer, with one path for each: security foundations, how LLM apps break, the OWASP LLM and agentic lists and MITRE ATLAS, layered defences against prompt injection, agent, tool and MCP security, the data and model supply chain, red-teaming and security evals, monitoring and incidents, and a route into the work. It suits developers and security learners who can already build. By the end you have a practice app defended or assessed end to end.

• AI
• Security
• Testing
• Career

An eight-stage path into AI governance, with a technical edition and a policy and compliance edition: the EU AI Act, NIST AI RMF, ISO/IEC 42001 and regional AI policies, an AI inventory with risk tiers and owners, data foundations and data protection, model assurance, human oversight, compliance and audit, and a route into the field. It suits engineers and data people as well as law, risk, audit and business graduates. By the end you can produce the documents governance depends on, with dated sources.

• AI
• Career

The revision companion to Sefism's AI agents and MCP roadmap: ten chapters on the agent loop, tool calling, structured output, memory, multi-agent patterns, frameworks, building MCP servers, evals, guardrails and shipping. It comes in Python and TypeScript editions with working snippets in each. It suits developers who already call model APIs and want to revise a concept quickly, copy a working shape or catch the classic mistake before users do.

• AI
• Programming
• Web

A quick-revision book on building features with LLM APIs: tokens, prompts, streaming, structured output, tool calling, embeddings and RAG, caching, batching, cost, evals and security, in 8 chapters and 56 concepts. It comes in OpenAI, Anthropic and Gemini editions, each showing the same ideas with that provider's official Python SDK. It suits engineers who can already call an HTTP API in Python and now need their AI features to be reliable, affordable and safe.

• AI
• Backend
• Programming

A 10-stage path for developers who can already call a model API and want to build agents that hold up in real use, in Python or TypeScript. It covers the agent loop by hand, structured output, memory and state, multi-agent patterns, LangGraph and the OpenAI and Claude agent SDKs, building and publishing an MCP server, evals, guardrails against prompt injection, and tracing and shipping. You finish with an evaluated, limited, traced agent and an MCP server a reviewer can try.

• AI
• Programming
• Web
• Backend

A quick-revision book on the Model Context Protocol for engineers building AI apps: hosts, clients and servers, JSON-RPC messages and lifecycle, transports, tools, resources and prompts, client features, OAuth for remote servers, building and testing servers with the official SDKs, security and tool design. It reflects the 2026-07-28 revision of the specification and notes what older revisions did differently. Use it while building or reviewing an MCP server, alongside the AI agents and MCP roadmap.

• AI
• Backend

Free, short, carefully drawn videos that explain statistics and machine learning one small step at a time: p-values, regression, PCA, random forests, gradient descent, neural networks and transformers. When a textbook formula refuses to make sense, a StatQuest video is usually the quickest way to the intuition behind it.

• AI
• Machine learning

Project Jupyter's free, open source environment for notebooks, code and data. You run code a cell at a time, see tables and charts right under it, and keep your reasoning in Markdown beside the code, which is how most data analysis is explored and shared. Installs with one pip command.

• AI
• Programming
• Machine learning

The standard first textbook for statistical learning, free as a PDF in two editions: ISLR2 with R labs and ISLP with Python labs. It explains regression, classification, resampling, regularisation, trees, SVMs, deep learning and unsupervised learning with the minimum of maths, and ends every chapter with a hands-on lab.

• AI
• Machine learning
• Programming

A seven-stage path for developers who can already ship a web or backend app and want to build products on large language models, in Python or TypeScript: tokens and context, prompting and model APIs, RAG with citations, evals in CI, tool calling and MCP servers, agents with durable execution, and running AI features in production. One project grows through every stage. By the end you can build, evaluate and run an LLM feature with retrieval, tools and agents, and explain what it costs, how you measured it and where it fails.

• AI
• Programming
• Web
• Backend

A quick-revision companion to the AI engineering roadmap in eight chapters: how LLMs work, prompting and model APIs, RAG, evals, tool calling and structured output, MCP, agents and durable execution, and running AI systems in production. It comes in Python and TypeScript editions, with snippets using the official SDKs, Pydantic or Zod. Skim a chapter after each roadmap stage, before changing a prompt or model, or before a technical conversation about AI features.

• AI
• Programming
• Web

Talks and full workshops from the AI Engineer conferences, including the AI Engineer World's Fair, where engineers, founders and researchers explain how they build with large language models and agents. Tracks cover agents, evals, context engineering, memory, security, platform engineering and more. It suits developers already building with LLMs who want to learn from teams shipping them in production.

• AI

The official n8n forum, where users ask and answer questions about workflows, nodes, expressions, self-hosting and errors, and share what they build. Searching it first solves most error messages, and answering questions there is a practical way to build skill and visibility.

• AI
• Career

A path for programmers and students who want to train models on data rather than only call an API: Python for data, the maths you need, data cleaning and features, classical machine learning, honest evaluation, deep learning in PyTorch and production MLOps, with an optional computer vision or NLP stage. By the end you can clean a real dataset, choose and evaluate a classical or deep learning model, and ship it as a reproducible, monitored service. The general focus is about 425 hours; vision or NLP adds a stage.

• AI
• Machine learning
• Programming

The official channel of DeepLearning.AI, founded by Andrew Ng, with lectures from its Coursera programmes including the Deep Learning Specialization and the Machine Learning Specialization, plus recorded events and conversations. The lectures are calm, structured and strong on intuition. It suits beginners and intermediate learners who want a trusted, course-like grounding in machine learning and deep learning.

• AI
• Machine learning

Gemini Notebook, called NotebookLM until July 2026, is Google's research notebook that answers questions using only the sources you add, such as PDFs, slides, notes, websites and videos, with citations back to the passage used. Students use it to study course material and turn it into study guides, quizzes, flashcards and audio or video overviews. It is free with daily limits, and higher limits come with Google AI plans.

• AI

A workflow automation platform, similar in spirit to Zapier, that you can self-host for free or use as a paid cloud service. You connect apps and APIs on a visual canvas, drop into JavaScript or Python when needed, and build AI agent workflows with language models and tools. It suits developers and technical students who want powerful automation with full control over their data and costs.

• AI
• DevOps
• Programming
• Web

GitHub Copilot is GitHub's AI coding assistant. It completes code as you type, answers questions in chat, runs as an agent that edits files and runs commands in your editor, and can take on issues as a cloud coding agent that opens pull requests. It works inside VS Code, Visual Studio, JetBrains IDEs, Xcode, Neovim and more. There is a free tier, and verified students get the Copilot Student plan at no cost.

• AI
• Dev tools

Cursor is an AI code editor built on VS Code, with agents that read your whole repository, plan changes across many files, run commands and review pull requests. It suits developers and students who want AI help inside their editor rather than a separate chat window. The free Hobby plan has limited agent use; serious daily use needs the $20 Pro plan or higher.

• AI
• Dev tools

A short, text-based course on writing prompts for everyday engineering work: how language models behave, the anatomy of an effective prompt, and using prompts to understand code, write code, refine answers, learn new topics and automate routine tasks. It suits developers and students who already use ChatGPT or a similar assistant and want clearer, more consistent answers after an evening or two of focused reading.

• AI

Chip Huyen's 2022 O'Reilly book on everything around the model: business framing, data engineering, training data, feature engineering, evaluation, deployment, distribution shift, monitoring, continual learning and MLOps infrastructure. It explains why models that score well in a notebook fail months after launch. It is for ML engineers, data scientists and engineers moving into ML who need to build systems that keep working in production.

• AI
• Machine learning
• System design
• DevOps

A free nine-week course from DataTalks.Club on putting machine learning models into production: experiment tracking with MLflow, orchestration, batch and streaming deployment, monitoring with Evidently and Grafana, and engineering practice such as testing, CI/CD and Terraform. It ends with an end-to-end project. It is for data scientists and engineers who can train a model but have never shipped or operated one.

• AI
• DevOps
• Machine learning

A modern deep learning textbook by Simon J.D. Prince, published by MIT Press in 2023 and free to download as a PDF. It explains how networks, transformers, diffusion models and reinforcement learning actually work, with clear figures and Python notebooks for every chapter. It suits students and engineers who can run a model but want to understand why it works, without a dense research-level text.

• AI
• Machine learning