Free AI Courses
Hand-picked from Anthropic, Google, Microsoft, OpenAI, DeepLearning.AI & more — ranked by trend
60 courses
Model Context Protocol (MCP) — Build Rich-Context AI Apps
Learn to build applications using the Model Context Protocol — connect Claude to tools, databases, and real-world data sources. The hottest skill in AI right now.
MCP: Build Rich-Context AI Apps with Anthropic
Learn the Model Context Protocol with Andrew Ng and Anthropic. Build agents that connect to live tools and data — the skill every AI developer needs in 2025.
Prompt Engineering Interactive Tutorial
Master prompt engineering with Claude through hands-on exercises. Covers chain-of-thought, few-shot prompting, avoiding hallucinations, and production best practices.
Hugging Face Agents Course
Build AI agents using the smolagents framework. Covers tool use, memory, multi-agent systems, and deploying agents to Hugging Face Spaces — with a certificate.
AI Agents in LangGraph
Build production-ready AI agents using LangGraph. Covers state machines, tool use, human-in-the-loop, and multi-agent coordination patterns.
Building Agentic RAG with LlamaIndex
Build advanced retrieval-augmented generation systems where AI agents decide what to retrieve and when. Covers routers, query engines, and multi-document agents.
Claude API Quickstart
Get up and running with the Claude API in minutes. Build your first AI-powered app with authentication, message streaming, and tool use.
Finetuning Large Language Models
Learn when and how to fine-tune LLMs vs. prompt engineering. Covers instruction tuning, LoRA, data preparation, and evaluation — with hands-on code.
Fine-Tune LLMs with Hugging Face
Hands-on guide to fine-tuning open-source LLMs like Llama and Mistral with LoRA/QLoRA. Covers data formatting, training, evaluation, and pushing to the Hub.
ChatGPT Prompt Engineering for Developers
The original prompt engineering course by OpenAI and DeepLearning.AI. Covers iterative prompting, summarizing, inferring, transforming, and building a chatbot.
Gemini API Quickstart
Start building with Gemini — Google's most capable AI model. Covers multimodal inputs, function calling, streaming, and the latest Gemini 2.0 features.
LangChain for LLM Application Development
Build complete LLM applications with LangChain. Covers chains, memory, agents, and RAG — the most popular LLM framework for production apps.
Introduction to Generative AI Learning Path
Google's official path to generative AI fundamentals. Covers LLMs, image generation, responsible AI, and building with Vertex AI — all with badges.
Building Systems with the ChatGPT API
Learn to build multi-step LLM pipelines and automated systems. Covers chaining calls, moderation, evaluation, and end-to-end app architecture.
Generative AI for Beginners (18 Lessons)
18-lesson curriculum from Microsoft covering everything from LLM basics to building RAG apps and AI agents. Hands-on with Azure OpenAI and open-source models.
OpenAI Cookbook
Practical recipes and code examples for building with the OpenAI API. Covers RAG, function calling, fine-tuning, embeddings, vision, and production patterns.
Hugging Face NLP Course
The comprehensive guide to NLP with Transformers. Covers text classification, NER, translation, summarization, and fine-tuning BERT/GPT models on custom data.
Practical Deep Learning for Coders
Jeremy Howard's legendary top-down deep learning course. Start building real models on day one — covers vision, NLP, tabular data, and diffusion models with PyTorch.
AI For Everyone
Andrew Ng's non-technical guide to AI for leaders, managers, and product teams. Understand what AI can and cannot do, and how to build an AI strategy.
Machine Learning Crash Course
Google's fast-paced introduction to machine learning with TensorFlow. Covers regression, classification, neural networks, and ML fairness with 25+ exercises.
Azure AI Fundamentals: Generative AI
Official Microsoft certification prep for AI-900. Covers Azure AI services, responsible AI principles, and building with Azure OpenAI — earn a free badge.
Discovering Faster Matrix Multiplication Algorithms with Reinforcement Learning (AlphaTensor)
Learn how reinforcement learning can be applied to discover novel and efficient matrix multiplication algorithms, as demonstrated by DeepMind's AlphaTensor. Understand the intersection of machine learning and computational mathematics to optimize one of the most fundamental operations in computer science.
Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning
Learn how to train a quadruped robot (ANYmal) to walk using massively parallel deep reinforcement learning techniques. Understand the computational strategies and algorithms that enable robots to learn complex locomotion tasks efficiently.
Magnetic Control of Tokamak Plasmas Through Deep Reinforcement Learning
Learn how to apply deep reinforcement learning techniques to control magnetic fields in tokamak plasma systems. Understand the intersection of machine learning and plasma physics through practical implementation of DRL algorithms for real-time plasma stabilization.
Outracing Champion Gran Turismo Drivers with Deep Reinforcement Learning
Learn how deep reinforcement learning can be applied to master complex racing simulations by training an AI agent to outperform professional Gran Turismo drivers. This project demonstrates advanced RL techniques including policy networks and training strategies used to achieve superhuman performance in competitive gaming environments.
BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning
Learn how to implement BC-Z, a behavioral cloning approach that enables robots to generalize to new tasks without task-specific training. This project demonstrates techniques for achieving zero-shot task adaptation in robotic systems through imitation learning principles.
Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild
Learners will understand how to train quadrupedal robots to navigate complex real-world environments using perceptive sensors and machine learning techniques. This course covers the methods for developing robust locomotion controllers that can adapt to diverse terrain and conditions without extensive manual engineering.
PaLI: A Jointly-Scaled Multilingual Language-Image Model
Learners will understand the architecture and design of PaLI, a state-of-the-art model that jointly scales language and vision capabilities across multiple languages. This resource provides insights into building unified multimodal systems that can process and generate content in diverse languages.
Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks
Learners will understand BEiT (BERT pre-training of Image Transformers) and how to leverage it for various vision and vision-language tasks. This course covers the methodology of treating images as a foreign language and how to apply pre-trained models to downstream applications.
VLMo: Unified Vision-Language Pre-Training with Mixture-of-Modality-Experts
Learn about VLMo, a unified pre-training approach that combines vision and language modalities using mixture-of-modality-experts architecture. Understand how to leverage multimodal expert routing and transformer-based models for improved vision-language understanding tasks.
CoCa: Contrastive Captioners are Image-Text Foundation Models
Learners will understand how CoCa combines contrastive learning with image captioning to create a unified foundation model for vision-language tasks. This resource provides insights into building and fine-tuning multimodal models that effectively represent both visual and textual information.
A Generalist Agent (Gato)
Learn about Gato, a generalist agent capable of handling diverse tasks across different modalities including text, images, and control problems. Understand the architecture and training approach that enables a single model to perform effectively across multiple domains without task-specific fine-tuning.
Flamingo: a Visual Language Model for Few-Shot Learning
Learn about Flamingo, a visual language model designed to perform few-shot learning tasks by combining vision and language understanding. This resource explores how to leverage multimodal models for efficient learning with minimal labeled examples.
Winoground: Probing Vision and Language Models for Visio-Linguistic Understanding
Learn how to probe and evaluate vision-language models for their ability to understand complex visio-linguistic relationships and grounded language understanding. Understand the limitations and capabilities of multimodal models through systematic testing and analysis.
VL-Adapter: Parameter-Efficient Transfer Learning for Vision-and-Language Tasks
Learn how to implement parameter-efficient transfer learning techniques for vision-and-language tasks using adapter modules. This project demonstrates methods to adapt pre-trained vision-language models to downstream tasks with minimal trainable parameters.
data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language
Learn about data2vec, a unified self-supervised learning framework that applies the same learning algorithm across speech, vision, and language modalities. Understand how to leverage unlabeled data to pretrain models that can be effectively fine-tuned for downstream tasks across multiple domains.
BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding
Learn about BLIP, a unified framework for vision-language pre-training that bootstraps language-image understanding. Understand how to leverage multimodal learning for tasks like image captioning and visual question answering.
Retroformer: Retrospective Large Language Agents with Policy Gradient Optimization
Learn how to implement Retroformer, a framework that uses policy gradient optimization to improve large language model agents through retrospective learning. This course covers advanced techniques for optimizing LLM agent behavior using reinforcement learning principles.
Faster sorting algorithms discovered using deep reinforcement learning
Learners will understand how deep reinforcement learning can be applied to discover novel and faster sorting algorithms beyond traditional hand-crafted solutions. This project demonstrates how AI can optimize fundamental computer science problems and achieve practical improvements in algorithm efficiency.
Direct Preference Optimization: Language Models as Reward Models
Learn Direct Preference Optimization (DPO), a technique that simplifies language model alignment by eliminating the need for separate reward models. Understand how to directly optimize language models using human preference data while maintaining computational efficiency.
Reward Design with Language Models
Learners will understand how to design effective reward functions for language models to guide their behavior and outputs. This course covers techniques for training language models using reward signals and optimizing their performance through careful reward design.
Grounding Large Language Models in Interactive Environments with Online RL
Learn how to ground large language models in interactive environments using online reinforcement learning techniques. Understand methods for enabling LLMs to learn and adapt through real-time interaction and feedback.
SeamlessM4T: Massively Multilingual & Multimodal Machine Translation
Learn about SeamlessM4T, a massively multilingual and multimodal machine translation system that enables translation across multiple languages and modalities. Understand how to implement and work with state-of-the-art translation models that handle both text and speech inputs.
Efficient Online Reinforcement Learning with Offline Data
Learn how to combine offline and online reinforcement learning to train agents more efficiently using pre-collected datasets. This course teaches methods for leveraging existing data while adapting to new environments through online interaction.
Mastering Diverse Domains through World Models (DreamerV3)
Learn how to build and train world models that can master diverse domains through the DreamerV3 framework. Understand how agents can learn generalizable representations and policies across different environments using latent world models and imagination-based planning.
Meta-Transformer: A Unified Framework for Multimodal Learning
Learn how Meta-Transformer provides a unified architecture for processing multiple modalities of data including images, text, audio, and video. Understand the framework's approach to multimodal learning and how to apply transformer-based models across different data types in machine learning applications.
Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning
Learners will understand the architecture and training methodologies behind CM3Leon, a state-of-the-art autoregressive multi-modal model that processes both text and images. This course covers pretraining techniques and instruction tuning strategies for scaling large multimodal models to handle complex vision-language tasks.
ImageBind: One Embedding Space To Bind Them All
Learn how ImageBind creates a unified embedding space that binds together multiple modalities including images, text, audio, depth, and video using a single neural network model. Understand the architecture and techniques that enable cross-modal alignment and retrieval tasks across different data types.
AudioGPT: Understanding and Generating Speech, Music, Sound, and Talking Head
Learn how to build an AI system that understands and generates various audio modalities including speech, music, sound effects, and talking head videos. Gain practical knowledge of multimodal audio processing and generation techniques using GPT-based architectures.
PaLM-E: An Embodied Multimodal Language Model
Learn about PaLM-E, a foundation model that combines language understanding with embodied reasoning for robotic control and multimodal tasks. Understand how to integrate visual perception with language models to enable agents to reason about and perform physical world interactions.
Aligning Perception with Language Models (Kosmos-1)
Learn how to align visual perception capabilities with large language models through the Kosmos-1 framework, which extends LLMs beyond text to handle multimodal inputs. Understand the key techniques for integrating vision and language understanding to create models that can process both images and text coherently.
LLM-RL-Visualized
A comprehensive visual reference guide containing 100+ algorithm maps for understanding Large Language Models and Reinforcement Learning concepts. Learners will gain visual insights into complex LLM and RL algorithms through structured, illustrated maps.
LLM-RL-Visualized
A comprehensive visual guide containing 100+ algorithm maps for understanding Large Language Models and Reinforcement Learning concepts. Learners will gain visual insights into how LLM and RL algorithms work and their relationships through curated algorithmic visualizations.
GPT in 60 Lines of NumPy
Learn to build a minimal GPT implementation using only NumPy in a compact 60-line script. Gain a deep understanding of how transformer-based language models work by implementing core concepts like attention mechanisms and token embeddings from scratch.
Welcome to the Big Model Era: Techniques and Systems to Train and Serve Bigger Models
Learn techniques and systems infrastructure for training and deploying large-scale models in the modern era of big models. Understand the architectural considerations, optimization strategies, and practical approaches needed to work effectively with large language models and massive neural networks.
Foundational Robustness of Foundation Models
Learn about the robustness properties and vulnerabilities of large foundation models in modern machine learning. Understand evaluation techniques and methodologies for assessing how foundation models handle adversarial inputs and distribution shifts.
femtoGPT: Pure Rust Implementation of a Minimal Generative Pretrained Transformer
Learn how to build a minimal Generative Pretrained Transformer (GPT) from scratch using pure Rust. Understand the core architecture and mechanics of transformer models used in modern language AI systems.
Byte Pair Encoding (BPE) Algorithm Implementation
Learn how to implement the Byte Pair Encoding algorithm with minimal, clean code commonly used in large language model tokenization. Understand the core mechanism behind how LLMs convert text into tokens for processing.
Build GPT from Scratch in Code
Learners will build a GPT language model from first principles, understanding each component of the architecture through hands-on code implementation. This course provides a detailed walkthrough of how transformer-based models work at a fundamental level.
A Visual Guide to Mamba and State Space Models
Learn about Mamba and state space models through visual explanations and intuitive examples. Understand how these modern architectures work and their applications in sequence modeling tasks.