AI Learning Resources

Andrew Ng's non-technical intro to AI for business leaders and curious minds. Learn what AI can and cannot do, how to navigate AI strategy, and ethical considerations.

Free online course from the University of Helsinki introducing AI concepts to non-technical audiences. 1 million+ students enrolled. Covers AI basics, machine learning, and societal impacts.

Comprehensive guide covering all major prompt engineering techniques: zero-shot, few-shot, chain-of-thought, tree-of-thought, and advanced prompting strategies for LLMs.

Cohere's free curriculum for learning large language models from basics to advanced RAG, fine-tuning, and deployment. Includes hands-on labs with the Cohere API.

Google's official generative AI learning path covering LLMs, image generation, responsible AI, and building generative AI applications with Google Cloud tools.

Top-down, code-first approach to deep learning using PyTorch and fastai. Highly recommended for developers who want to build real-world deep learning applications quickly.

Hands-on NLP with the Transformers library. Learn to fine-tune BERT, GPT-2, and other models. Covers tokenization, training, evaluation, and model deployment.

5-course series covering neural networks, hyperparameter tuning, regularization, CNNs, RNNs, and sequence models. By Andrew Ng - the definitive deep learning curriculum.

An 8-week course on AI alignment and safety research. Covers value alignment, robustness, interpretability, and governance for those interested in AI safety careers.

MIT's intensive deep learning short course. Covers foundations, CNNs, RNNs, GANs, reinforcement learning, and research frontiers. Lecture videos freely available.

Short hands-on course teaching LangChain fundamentals: chains, memory, agents, and document Q&A. Co-taught by LangChain founder Harrison Chase.

Stanford's flagship ML course with full lecture videos online. Covers linear algebra foundations, supervised learning, unsupervised learning, and theoretical ML concepts.

4-course specialization on deploying ML systems in production. Covers concept/data drift, CI/CD for ML, model monitoring, and ML system design by Andrew Ng.

Practical short course on building multi-step LLM systems: chains of prompts, input evaluation, output checking, and end-to-end customer service applications.

Learn advanced RAG techniques including sentence window retrieval, auto-merging retrieval, and RAG triad evaluation. Hands-on with LlamaIndex and TruEra.

Stanford's advanced NLP course covering word vectors, neural dependency parsing, language models, transformers, and the latest large language model architectures.

Intensive bootcamp on building LLM-powered apps covering prompt engineering, LLM APIs, fine-tuning, deployment, and UX for AI products. Used by thousands of engineers.

Beautiful visual explainer of neural networks, backpropagation, and gradient descent. The best mathematical intuition builder available - essential watching before any ML course.

Josh Starmer breaks down statistics and ML concepts with unmatched clarity. Best series for understanding the math behind ML without a PhD. Covers SVMs, random forests, PCA, neural nets.

Accessible explanations of the latest AI developments, models, and research papers. Ideal for staying current with GPT-4, Claude, Gemini, and frontier AI developments without deep technical background.

University of Nottingham's Computerphile channel explains AI, ML, and deep learning concepts clearly. Great for intuition on neural networks, NLP, and AI ethics without heavy math.

Harrison Kinsley's hands-on Python ML series covering scikit-learn, TensorFlow, and neural networks. Practical, project-based approach covering clustering, regression, classification, and deep learning.

Károly Zsolnai-Fehér summarizes the latest AI research papers in 2-5 minutes with enthusiasm and clarity. Essential for staying current with computer vision, RL, and generative AI research.

Deep long-form interviews with the world's leading AI researchers and practitioners including Sam Altman, Yann LeCun, Geoffrey Hinton, Andrej Karpathy, and Demis Hassabis.

Practical MLOps and deep learning content from W&B. Covers experiment tracking, model evaluation, LLM fine-tuning, and hands-on tutorials with PyTorch and Keras.

Detailed walkthroughs of landmark AI research papers including GPT-4, DALL-E, AlphaFold, and RLHF. Best channel for understanding how frontier AI systems actually work at the research level.

Stanford's legendary computer vision course lectures on YouTube. Covers CNNs, object detection, image segmentation, generative models (GANs/VAEs), and video understanding.

Official Hugging Face YouTube with tutorials on Transformers, fine-tuning, diffusion models, and PEFT techniques. Directly from the team building the most widely used AI library.

Official DeepMind channel featuring research presentations on AlphaFold, Gemini, reinforcement learning, AI safety, and neuroscience-inspired AI from the world's leading research lab.

Practical crash course on building AI-powered applications with Python including ChatGPT API integration, LangChain chatbots, image generation with DALL-E, and Whisper transcription.

Jeff Delaney's lightning-fast explainers on AI tools, frameworks, and concepts. Perfect for developers who want a quick mental model of GPT, diffusion models, and AI coding tools.

Foundation-level AI/ML certification from AWS. Validates understanding of AI concepts, AWS AI services (SageMaker, Rekognition, Comprehend), and responsible AI practices.

Entry-level Azure AI certification. Covers AI workloads, machine learning concepts, computer vision, NLP, and responsible AI principles on Azure.

6-course professional certificate covering machine learning, deep learning with Keras/TensorFlow/PyTorch, computer vision, and deploying AI models. Highly regarded for career transitions.

Validate your TensorFlow skills building and training neural network models. Covers image classification, NLP with TensorFlow, time series, and sequence models. Taken via remote proctored exam.

Associate-level certification for designing and implementing Azure AI solutions. Covers Azure Cognitive Services, Azure ML, conversational AI, and knowledge mining.

Specialty certification validating ability to design, implement, deploy and maintain ML solutions on AWS. Covers data engineering, exploratory analysis, modeling, and MLOps with SageMaker.

Professional certification for ML engineers working on Google Cloud Platform. Covers ML problem framing, data preparation, model development, and MLOps on Vertex AI.

Professional certification for ML practitioners using Databricks lakehouse platform. Covers MLflow, feature engineering, model deployment, and ML pipelines on Apache Spark.

NVIDIA's instructor-led certification in deep learning fundamentals. Hands-on training with GPU-accelerated deep learning covering CNNs, data augmentation, transfer learning, and deployment.

10-course program covering data science tools, Python for data science, SQL, data visualization, machine learning, and a capstone project. One of the most popular certificates on Coursera.

Vendor-neutral AI certification covering ML model development, deep learning, NLP, computer vision, AI ethics, and deployment. Recognized by CompTIA as complementary to IT certifications.

Harvard's free AI course covering search algorithms, knowledge representation, uncertainty, optimization, machine learning, neural networks, and NLP. Certificate available via edX.

Deep dive into generative AI and LLMs covering transformer architecture, fine-tuning, RLHF, and deployment considerations. Co-developed by AWS and taught by Deeplearning.AI.

The most practical and widely recommended ML book. Covers end-to-end ML projects, classification, regression, SVMs, decision trees, ensemble methods, and deep learning with TensorFlow/Keras. Now in 3rd edition.

The definitive academic textbook on deep learning by its founders. Covers probability theory, information theory, linear algebra, optimization, CNNs, RNNs, autoencoders, and generative models. Free online.

Remarkably concise book covering the essentials of ML in ~130 pages. Covers supervised learning, neural networks, regularization, SVMs, and ensembles. Excellent refresher or introduction for experienced engineers.

Former Google China president Kai-Fu Lee analyzes the US-China AI race, predicts which jobs will be automated, and explores AI's societal and economic implications. A must-read for executives and policymakers.

Compelling narrative account of the AI alignment problem - the challenge of getting AI systems to do what we actually want. Covers value alignment, fairness, interpretability, and AI safety research.

UC Berkeley AI professor Stuart Russell reframes the AI safety problem and proposes a new approach to building provably beneficial AI. Essential reading for anyone thinking seriously about AGI risks.

The gold standard graduate textbook for probabilistic machine learning. Covers Bayesian methods, kernel machines, graphical models, EM algorithm, and mixture models. Free PDF available from Microsoft Research.

Covers the mathematical foundations necessary for modern ML: linear algebra, analytic geometry, matrix decompositions, probability, optimization, and dimensionality reduction. Free PDF from Cambridge.

The definitive practical guide to NLP with Hugging Face Transformers. Covers text classification, named entity recognition, question answering, summarization, and generation. Written by the Hugging Face team.

Comprehensive guide to designing production ML systems from a systems engineering perspective. Covers data engineering, feature engineering, model development, deployment, monitoring, and MLOps at scale.

DeepMind co-founder Mustafa Suleyman argues that AI and synthetic biology represent an unprecedented technological wave that will reshape power, democracy, and human civilization. Essential for policymakers.

MIT physicist Max Tegmark explores near-future AI scenarios, existential risks, and what it means to flourish in a world with superintelligent AI. Accessible and thought-provoking for non-technical readers.

Practical guide to building production-ready LLM-powered applications with LangChain, LlamaIndex, vector databases, and OpenAI APIs. Covers chatbots, RAG systems, and AI agents with Python.

Jay Alammar's visual and intuitive explanation of the Transformer architecture. The best resource for understanding attention mechanisms, positional encoding, and encoder-decoder structure.

Visual explanation of BERT, ELMo, and ULMFiT - the pre-training approaches that revolutionized NLP. Explains context-dependent word embeddings, masked language modeling, and next-sentence prediction.

Interactive browser tool to visualize how neural networks learn. Experiment with hidden layers, activation functions, and datasets in real time. The best hands-on intro to neural network intuition.

Gorgeous interactive visual essay explaining decision trees, overfitting, and machine learning fundamentals. Animated scrollytelling format makes abstract concepts immediately concrete and memorable.

Peer-reviewed journal publishing interactive visual explanations of ML research. Articles on attention, feature visualization, circuits in neural nets, and multimodal neurons. Research-grade but beautifully explained.

Line-by-line implementation of the Transformer architecture from 'Attention Is All You Need' in PyTorch with detailed annotations. The definitive code tutorial for understanding transformers from first principles.

OpenAI Head of Safety's comprehensive technical blog. Definitive deep-dive articles on attention mechanisms, policy gradients, diffusion models, LLM alignment, and agent frameworks. Extremely well-written and thorough.

Former Tesla AI Director's influential technical blog. Classic posts include 'The Unreasonable Effectiveness of RNNs', 'A Recipe for Training Neural Networks', and 'Hacker's Guide to Neural Networks'.

Clear, concise documentation on ML concepts including linear regression, logistic regression, gradient descent, neural networks, and evaluation metrics. Great quick reference and study guide.

The companion book to fast.ai's course, available free online. Covers deep learning from foundations to state-of-the-art tabular, NLP, and computer vision applications with Jupyter notebooks.

Official OpenAI repository of practical examples and guides for using the OpenAI API. Covers embeddings, fine-tuning, function calling, batch processing, and building production-ready AI applications.

Official Hugging Face tutorials covering fine-tuning LLMs with PEFT/LoRA, diffusion models, text generation, translation, and building inference pipelines. Regularly updated with latest models.

Google's free self-study guide to ML basics with interactive visualizations, videos, and programming exercises. Covers framing ML problems, loss reduction, generalization, and neural networks. 15+ hours of content.

Kaggle's free interactive micro-courses covering Python, pandas, ML fundamentals, deep learning, NLP, computer vision, and intro to AI. Browser-based notebooks, instant feedback, and certificates. Used by millions of data scientists worldwide.

Build production-ready AI agents using LangGraph — the stateful agent orchestration framework. Covers ReAct agents, tool use, memory, human-in-the-loop, and multi-agent collaboration patterns. Co-taught with the LangChain team.

Step-by-step guide to building a GPT-class LLM entirely from scratch in Python — from tokenization and attention mechanisms to pre-training and fine-tuning with RLHF and instruction tuning. The most hands-on LLM book of 2024.

Hugging Face's free hands-on course on deep reinforcement learning covering Q-Learning, Deep Q-Networks (DQN), Policy Gradient methods, Proximal Policy Optimization (PPO), and multi-agent RL. Includes interactive Unity environments and publishes trained agents to the Hugging Face Hub. Used by 70,000+ learners.

Amazon's official free ML foundations learning path on AWS Skill Builder. Covers ML concepts, AWS AI/ML services (SageMaker, Rekognition, Comprehend, Bedrock), and hands-on labs. Essential preparation for the AWS AI Practitioner and ML Specialty certifications. Used by hundreds of thousands of AWS practitioners worldwide.

Andrej Karpathy's widely shared 1-hour lecture giving a clear, no-fluff mental model of what LLMs are and how they work — covering tokenization, transformer architecture, pretraining, fine-tuning, RLHF, emergent capabilities, and practical LLM system security risks. Considered one of the best single-video introductions to LLMs for engineers who already know the basics of ML. Viewed 3M+ times.

OpenAI's comprehensive educational resource for deep reinforcement learning. Covers the fundamentals of RL theory, key algorithms (VPG, TRPO, PPO, DDPG, TD3, SAC), and hands-on implementations in PyTorch. Includes a curated reading list of landmark papers and documentation on running your own RL experiments. The gold-standard starting point for serious RL research and RLHF practitioners.

Google's official responsible AI documentation and best practices, covering fairness, interpretability, privacy, security, and safety in ML systems. Includes practical guidance on identifying and mitigating bias in training data, model interpretability techniques, and checklists for responsible AI product development. Essential reference for teams building production AI systems that must meet ethical and regulatory standards.

Anthropic's official interactive tutorial for prompt engineering with Claude. Covers foundational techniques — clarity, role specification, chain-of-thought, XML structuring — through hands-on exercises directly in the browser. Essential starting point for anyone building applications with Claude or other large language models. Regularly updated to reflect Claude's latest capabilities.

Microsoft's open-source 12-week, 24-lesson curriculum covering the full scope of AI for beginners. Includes symbolic AI, neural networks, computer vision, NLP, generative AI, and responsible AI. All lessons include Jupyter notebooks with Python and TensorFlow/PyTorch exercises. Used by hundreds of thousands of learners on GitHub. Ideal for developers transitioning into AI.

DeepLearning.AI's ever-expanding catalogue of free 1–3 hour short courses co-developed with leading AI companies including OpenAI, Anthropic, Google, Mistral, Meta, and AWS. Topics span prompt engineering, RAG, fine-tuning, AI agents, LangChain, vector databases, multimodal AI, and AI safety. Each course is taught by practitioners from the sponsoring company and includes hands-on coding exercises.

Stanford's CS25 seminar series bringing together leading researchers to explain and extend the Transformer architecture. Lectures cover attention mechanisms, vision transformers (ViT), language model scaling, multimodal transformers, protein structure prediction (AlphaFold), and the latest research frontiers. Guest speakers have included researchers from DeepMind, OpenAI, Google Brain, and Meta AI. All lectures are freely available on YouTube and the course website. Essential for engineers who want to understand frontier model architectures at a research depth.

Google's official open-source cookbook for the Gemini API, with 100+ Jupyter notebook examples covering multimodal prompting, function calling, code execution, long-context document analysis, grounding with Google Search, embeddings, and Gemini integration with LangChain and LlamaIndex. Kept continuously updated as new Gemini capabilities launch. The fastest path to production-ready Gemini integrations for developers building with Google AI Studio or Vertex AI.

Simon Willison (co-creator of Django) maintains one of the most practical and technically rigorous independent blogs covering LLMs, AI tools, and AI safety. Posts cover hands-on experiments with every major model release, the LLM CLI tool (a Python CLI for running local and API-based LLMs), prompt injection attacks, retrieval-augmented generation, and critical analysis of AI company claims. Essential reading for developers who want opinionated, practitioner-level AI coverage without hype.

Anthropic's engineering guide to building reliable LLM agents in production. Distinguishes workflows from agents, walks through composable patterns — prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer loops — and argues for starting with the simplest approach that works before reaching for autonomous agents. Includes practical guidance on tool design, when agentic systems are worth the cost, and how to keep them debuggable.

Hands-on short course covering the full RAG pipeline: chunking and embedding documents, vector retrieval, reranking, and grounding LLM responses in retrieved context to reduce hallucination. Teaches practical evaluation of retrieval quality and answer faithfulness, with code you can adapt to build production question-answering systems over private knowledge bases.

Google's practical prompting handbook for getting useful results from Gemini across Workspace apps. Introduces a simple prompt formula — persona, task, context, and format — and provides role-based example prompts for marketing, sales, HR, executives, and customer service. A fast, approachable on-ramp for non-technical professionals adopting generative AI in everyday work.

The official documentation for the Model Context Protocol, an open standard for connecting AI assistants to external data sources, tools, and systems through a uniform client-server interface. Covers the protocol architecture, building MCP servers in Python and TypeScript, exposing tools, resources, and prompts, and integrating with clients like Claude Desktop and IDEs. Essential for developers building agentic systems that need standardized, reusable connections to real-world context instead of bespoke one-off integrations.

Andrej Karpathy's acclaimed free video course building neural networks from scratch in Python, culminating in a from-first-principles implementation of a GPT-style transformer (nanoGPT / makemore). Starts with micrograd and backpropagation by hand, then progressively constructs language models, explaining autograd, tokenization, attention, and training dynamics in depth. The single best resource for developers who want to truly understand how modern LLMs work under the hood rather than just calling an API.

Andrew Ng's updated flagship Machine Learning Specialization, the modern successor to his original ML course. Three courses covering supervised learning (regression, classification), advanced learning algorithms (neural networks, decision trees), and unsupervised learning, recommenders, and reinforcement learning. Uses Python, NumPy, scikit-learn, and TensorFlow with hands-on labs. The definitive foundational ML program for beginners who want rigorous intuition plus practical implementation before moving on to deep learning or LLMs.

Anthropic's free course teaching how to work effectively, efficiently, and responsibly with AI systems like Claude. Built around a practical framework (Delegation, Description, Discernment, and Diligence) for collaborating with AI, it covers how to structure prompts, evaluate outputs critically, and use generative AI ethically. A strong, vendor-grounded starting point for non-technical professionals and teams who want durable habits for working alongside AI rather than one-off prompt tricks.

A hands-on DeepLearning.AI course on building AI agents that plan, use tools, reflect, and coordinate to complete multi-step tasks. Covers core agentic design patterns — reflection, tool use, planning, and multi-agent collaboration — and how to compose them into reliable workflows with evaluation and guardrails. Aimed at developers who already understand LLM basics and want to move from single prompts to autonomous, tool-using agent systems that operate over multiple steps.

Google and Kaggle's free intensive program covering the foundations of generative AI in a compressed, project-driven format. Topics span foundation models and prompting, embeddings and vector databases, retrieval-augmented generation, AI agents, and building and evaluating LLM applications, with accompanying whitepapers and hands-on codelabs. A practical, fast-paced path for developers who want breadth across the modern generative-AI stack with runnable Kaggle notebooks.

Anthropic's practical guidance for building reliable AI agents with Claude, drawn from its widely cited 'Building Effective Agents' engineering work. Covers when to use simple workflows versus autonomous agents, common agent patterns (prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer), tool design and tool use, and how to keep agentic systems debuggable and cost-effective. A grounded, pattern-first path for developers moving from single prompts to multi-step, tool-using systems.

Hugging Face's free, hands-on course on building AI agents from the ground up. Covers agent fundamentals, the think-act-observe loop, tools and function calling, popular agent frameworks (smolagents, LlamaIndex, LangGraph), retrieval and memory, and evaluating agentic systems, culminating in a capstone project and certification. A practical route for developers who want to build and ship working agents with open-source tooling.

Google Cloud's business-focused certification validating a foundational, strategic understanding of generative AI. Covers core generative-AI and LLM concepts, real-world enterprise use cases, Google Cloud's AI stack (Vertex AI, Gemini), responsible AI, and how to identify and drive AI adoption across an organization. Aimed at leaders, product managers, and non-engineers who need to make informed decisions about deploying generative AI, rather than build models themselves. No prerequisites; 90 minutes, 50-60 multiple-choice questions, $99 plus applicable tax, valid for three years, sat remotely proctored or at a test centre. Vendor-specific by design — pair it with a vendor-neutral governance credential if the goal is regulatory competence rather than platform fluency.

Anthropic's open-source collection of runnable code recipes and notebooks for building with Claude. Covers practical patterns including tool use and function calling, retrieval-augmented generation, vision and PDF processing, structured outputs (JSON), prompt caching, sub-agents, and agentic workflows. A hands-on, example-first resource for developers who want working reference implementations rather than conceptual overviews.

A short course from DeepLearning.AI and Anthropic on the Model Context Protocol (MCP), the open standard for connecting AI applications and agents to external tools, data, and resources. Learn how MCP standardizes context sharing between LLM apps and services, how to build MCP servers and clients, and how to connect Claude and other models to real tools and data sources to build rich-context, interoperable AI applications.

Google's practical prompting course teaching a repeatable framework for writing effective prompts across generative-AI tools. Covers the task-context-references-evaluate-iterate approach, prompting for everyday work tasks like summarizing, drafting, and data analysis, using AI as a thought partner and role-playing assistant, and chaining prompts for multi-step workflows. Designed for professionals who want to get consistently better results from tools like Gemini and other LLMs without a technical background.

Microsoft's open-source, multi-lesson course on building AI agents, published on GitHub with runnable code examples. Covers agentic design patterns, tool use and function calling, agentic RAG, planning and multi-agent systems, metacognition, and production deployment concerns like trust and safety. Framework-inclusive with examples spanning Azure AI, Semantic Kernel, and AutoGen, it is a hands-on starting point for developers who want to build working agents rather than just read about them.

A short, free course from DeepLearning.AI and OpenAI teaching developers how to use large language models via the API to build applications. Covers prompting principles, writing clear and specific instructions, giving the model time to reason, iterative prompt development, and practical capabilities like summarizing, inferring, transforming, and expanding text, ending with building a simple chatbot. Taught by Isa Fulford and Andrew Ng with runnable Jupyter notebooks so developers can experiment as they learn.

LangChain's free course teaching how to build controllable, stateful agent applications with LangGraph. Covers graph-based agent architectures, state and memory, human-in-the-loop workflows, breakpoints and time travel for debugging, persistence, and deploying agents to production. Delivered with video lessons and runnable notebooks, it is a practical resource for developers who want reliable, production-grade agentic systems rather than brittle single-prompt chains.

The leading AI governance credential for compliance, legal, and privacy professionals. Covers AI laws and emerging regulation, the EU AI Act, GDPR intersections, liability and intellectual property, AI risk management, and responsible AI practice across development and deployment. Certification runs on two-year terms with continuing-education requirements.

Implementation-focused credential for building an AI management system to ISO/IEC 42001. Covers AI policy, cross-functional oversight, mandatory risk assessment and treatment, data quality and third-party supplier management, monitoring, and preparation for certification audit. Best suited to governance leads and implementation consultants rather than to legal interpretation.

Audit and assurance credential for assessing AI management systems against ISO/IEC 42001. Covers planning, leading, and reporting audits, including second-party auditing of suppliers and subcontractors. The strongest fit for internal and external auditors, certification bodies, and third-party risk assessors testing whether AI governance actually works.

Enterprise risk credential extending established GRC practice into AI. Covers AI models and frameworks, organisational processes, ownership and accountability, policy and training, regulatory compliance and legal considerations, and AI trustworthiness. Assumes an existing IT-risk background — the programme covers AI-specific risk and deliberately does not re-teach foundational IT risk.

Audit credential covering AI governance and risk, AI operations, and AI-specific auditing tools and techniques. Requires CISA or another qualifying designation as a prerequisite — it is an advanced credential, not an entry point. Best suited to internal and external auditors whose objective is AI assurance rather than regulatory interpretation.

Security-weighted AI governance credential covering AI programme management, AI risk management, and AI technologies and controls, including regulatory requirements, policy, data governance, and incident response. Requires CISM or CISSP as a prerequisite. Relevant to compliance work but centred on security rather than on broad AI governance.