AI Roadmaps

Learn AI from free YouTube videos, in a sensible order.

Every step is one video and, where it helps, one GitHub repo to code along with. No books, no paywalls, no playlists to sort through. Tick steps off as you go; your progress stays in your browser.

43 roadmaps · 540 videos · 269 hours · 165 repos · free, no account

Python & Math for AI13 steps · 6h 54m
  1. Python from zero2h 02m
  2. NumPy arrays58 min
  3. Pandas for data1h 34m
  4. Plotting with Matplotlib31 min
  5. Vectors10 min

Roadmaps

Recently added: AI for Finance (Nov 13) · World Models (Nov 9) · Self-Supervised & Contrastive Learning (Nov 8)

01
Beginner13 steps · 6h 54m

Python & Math for AI

The Python, NumPy, linear algebra, calculus, and probability you actually need before touching models.

02
Beginner14 steps · 5h 10m

Machine Learning

Classic ML the way practitioners use it: regression, trees, boosting, evaluation, and scikit-learn.

03
Intermediate14 steps · 11h 01m

Deep Learning

Neural networks from the inside: backprop, build your own autograd, CNNs, RNNs, and PyTorch.

04
Intermediate15 steps · 22h 04m

LLMs from Scratch

Understand and build GPT: tokenizers, attention, transformers, GPT-2, Llama, mixture of experts.

05
Beginner14 steps · 10h 07m

AI Engineering

Build real products on top of LLMs: prompting, APIs, structured output, RAG, evals, and local models.

06
Intermediate14 steps · 4h 56m

AI Agents & MCP

Agents that call tools, use MCP servers, plan, and coordinate, and how to tell if they work.

07
Advanced11 steps · 6h 35m

Fine-Tuning & Reasoning Models

Adapt open models with LoRA, align them with RLHF and DPO, and train reasoning with GRPO.

08
Intermediate12 steps · 4h 40m

Computer Vision

Teach machines to see: CNNs, transfer learning, YOLO detection, segmentation, ViT, and CLIP.

09
Intermediate12 steps · 9h 21m

Generative AI: Images & Video

How machines create images and video: VAEs, GANs, diffusion, latent diffusion, and flow matching.

10
Advanced15 steps · 6h 40m

Run, Serve & Scale AI

Get models into production and make them fast: Docker, MLOps, monitoring, quantization, vLLM, GPUs, and FSDP.

11
Intermediate11 steps · 3h 28m

Reinforcement Learning

Agents that learn by trial and error: from Bellman equations and Q-learning to PPO, AlphaGo, and the RL behind reasoning models.

12
Intermediate10 steps · 4h 07m

Speech & Audio AI

How machines hear and speak: waveforms, spectrograms, speech recognition with Whisper, text-to-speech, audio codecs, and voice agents.

13
Advanced10 steps · 7h 28m

Robotics & Embodied AI

Teach robots with data: robot learning basics, imitation learning, ACT, diffusion policies, vision-language-action models, and sim-to-real.

14
Advanced12 steps · 5h 36m

Interpretability

Open the black box: circuits, superposition, sparse autoencoders, induction heads, grokking, and tracing what LLMs compute.

15
Intermediate12 steps · 4h 17m

AI Safety & Alignment

Why powerful AI can go wrong and what researchers do about it: specification gaming, inner alignment, deception, oversight, and evals.

16
Intermediate12 steps · 3h 14m

Recommender Systems

How Netflix, YouTube, and TikTok pick what you see: collaborative filtering, matrix factorization, two-tower retrieval, ranking, and evaluation.

17
Intermediate11 steps · 3h 48m

Time Series & Forecasting

Predict what happens next: trend, seasonality, ARIMA, boosted trees, deep models, and today's pretrained forecasting models.

18
Advanced11 steps · 3h 26m

Graph Neural Networks

Learn from data shaped like networks: node embeddings, message passing, GCN, GAT, GraphSAGE, link prediction, and real uses.

19
Advanced12 steps · 11h 09m

Multimodal & Vision-Language Models

How models see and talk: CLIP, SigLIP, Flamingo, BLIP-2, LLaVA-style models, and building a vision-language model yourself.

20
Advanced11 steps · 11h 37m

GPU Programming & CUDA

Understand why GPUs are fast and write your own kernels: CUDA basics, memory, tiled matmul, Triton, and FlashAttention.

21
Intermediate15 steps · 4h 28m

Data Engineering for AI

Move, store, clean, and version the data that models learn from: SQL, Parquet, DuckDB, pipelines, labeling, and LLM-scale dataset curation.

22
Intermediate14 steps · 3h 06m

Math for ML, Deeper

The next layer of math behind ML papers: change of basis, SVD, Jacobians, Hessians, convexity, Lagrange multipliers, Gaussians, and KL divergence.

23
Advanced15 steps · 7h 04m

Bayesian ML & Probabilistic Programming

Reason with uncertainty: priors and posteriors, MCMC, PyMC, hierarchical models, Gaussian processes, variational inference, and Bayesian deep learning.

24
Advanced14 steps · 5h 27m

TinyML & Edge AI

Shrink models and run them on microcontrollers, phones, and small boards: pruning, quantization, distillation, TFLite Micro, Edge Impulse, ONNX, and ExecuTorch.

25
Intermediate12 steps · 5h 11m

AI for Science

How AI is changing biology and chemistry: protein folding, AlphaFold, protein language models, protein design, molecules in Python, and AI weather models.

26
Intermediate13 steps · 4h 27m

Knowledge Graphs & GraphRAG

Store facts as connected entities, query them with Cypher, build graphs from text with LLMs, and use them for retrieval that follows relationships.

27
Intermediate12 steps · 4h 58m

Synthetic Data & Data-Centric AI

Improve models by improving their data: find label errors, label less, augment, generate synthetic tables and LLM training sets, and know when synthetic data backfires.

28
Advanced12 steps · 3h 26m

Federated Learning & Privacy-Preserving ML

Train models without collecting the data: federated averaging, Flower simulations, differential privacy, secure aggregation, encryption, and the attacks these defend against.

29
Intermediate13 steps · 4h 13m

Causal Inference & Causal ML

Go from 'X is correlated with Y' to 'X causes Y': potential outcomes, causal graphs, matching, natural experiments, and machine learning for treatment effects.

30
Intermediate12 steps · 5h 00m

Evaluation & Benchmarking for LLMs

Measure language models and LLM apps properly: classic metrics, public benchmarks and leaderboards, LLM judges, error analysis, and open-source eval tools.

31
Advanced14 steps · 5h 51m

Quantum Machine Learning

Learn qubits, gates, and circuits, then train variational circuits, quantum classifiers, and quantum kernels, and see where the limits are.

32
Intermediate12 steps · 5h 27m

LLM Security & Red Teaming

Learn how LLM apps get attacked, from prompt injection and jailbreaks to data extraction and poisoning, then test and defend your own.

33
Intermediate12 steps · 6h 44m

Medical AI & Imaging

Work with medical images and clinical data: DICOM, segmentation with U-Net, MONAI, and nnU-Net, pathology, foundation models, bias, and regulation.

34
Advanced13 steps · 8h 00m

Self-Driving Cars & Autonomous Systems

Follow the self-driving stack from sensors and localization through perception, prediction, and planning to end-to-end models and simulation.

35
Advanced13 steps · 6h 59m

Mixture of Experts & Efficient LLM Architectures

See how modern LLMs grow without getting slower: sparse mixture of experts, cheaper attention, linear attention, state space models, and hybrids.

36
Advanced15 steps · 6h 08m

3D Vision, NeRFs & Gaussian Splatting

Turn photos into 3D: camera geometry, stereo, structure from motion, depth networks, neural radiance fields, Gaussian splatting, and feed-forward reconstruction.

37
Intermediate14 steps · 4h 43m

Explainable ML: SHAP, LIME & Model Debugging

Explain what your models do and why: glass-box models, feature importance, PDPs, LIME, SHAP and its limits, counterfactuals, saliency maps, and error analysis.

38
Intermediate15 steps · 4h 18m

Classic NLP Before LLMs

The NLP toolkit that came before transformers: tokenizing and normalizing text, TF-IDF, n-grams, Naive Bayes, HMM tagging, NER, topic models, word vectors, and seq2seq with attention.

39
Intermediate11 steps · 4h 41m

Anomaly Detection

Find the rare and the unusual: statistical outliers, distance and density methods, isolation forests, one-class SVMs, autoencoders, fraud with imbalanced data, and explaining why a point was flagged.

40
Intermediate10 steps · 3h 35m

Document AI & OCR

Get text, tables, and fields out of scans and PDFs: image cleanup, classic OCR engines, handwriting recognition, layout-aware models, PDF-to-Markdown tools, and vision-language models that read pages directly.

41
Advanced12 steps · 4h 52m

Self-Supervised & Contrastive Learning

How models learn strong features from unlabeled data: contrastive losses, SimCLR, BYOL, masked autoencoders, DINO through DINOv3, and JEPA's predict-the-representation idea.

42
Advanced10 steps · 6h 07m

World Models

Models that learn how an environment responds to actions, then plan or train agents inside it: model-based RL, Dreamer, MuZero, Genie, world foundation models, and JEPA-style planning.

43
Intermediate11 steps · 8h 26m

AI for Finance

Machine learning for markets and money: how quant firms work, portfolios and factors, honest backtesting, financial text, credit risk, and LLMs in finance. Education, not investment advice.

Pick a path

Repos worth starring

The repositories these roadmaps code along with.

Questions

How can I learn AI for free?

Pick a roadmap and watch its videos in order. All 43 roadmaps are made of free YouTube videos and open-source GitHub repos, with no sign-up, paywall, or trial.

Where should a complete beginner start?

Start with Python & Math for AI (6h 54m of video), then Machine Learning and Deep Learning. After that, choose a roadmap for what you want to build: LLMs, agents, computer vision, generative AI, or deployment.

How long does each roadmap take?

Between 3h 06m and 22h 04m of video, 269 hours in total. Plan for about twice that if you code along and build one of the projects at the end of each roadmap.

Do I need an account to track my progress?

No. Tick off each step as you finish it. Your progress is saved in your browser, and nothing is sent anywhere.

How are the videos chosen?

One specific video per step, never a whole playlist. Clear and short beats famous and long, every video is free to watch, and links are checked regularly so dead videos get replaced. Anyone can suggest a better one on GitHub.