After the roadmap
You know the fundamentals. Now go build.
Finishing the roadmap means you can read, write, and reason about AI. This is where you turn that into real skill — pick a track, ship projects, and keep leveling up. Every link here is free.
← Still working through the roadmap?Build & ship real projects
The fastest way to level up after the fundamentals is to build things end to end and put them in a portfolio.
Specialize — go deeper on a track
Pick a direction and go deep. These are the advanced areas people most often move into next.
- Computer Vision — Stanford CS231n Object detection, segmentation, and modern vision architectures. Free notes.
- NLP — Hugging Face NLP Course Transformers, tokenizers, and fine-tuning for language, hands-on and free.
- Reinforcement Learning — HF Deep RL Course Train agents that learn from reward. Free, project-based.
- Diffusion & image generation — HF Diffusion Course How Stable-Diffusion-style models actually work, and how to build them.
- LLMs from scratch — Karpathy: Zero to Hero Build a GPT by hand. The best deep dive into how LLMs work.
- Graph Neural Networks — Stanford CS224W Learning on graphs — recommendations, molecules, social networks.
- Audio & speech — HF Audio Course Speech recognition, text-to-speech, and audio classification, hands-on.
- Time series & forecasting — Forecasting: Principles & Practice The free, canonical book on predicting what happens next.
- Recommender systems — Google course How the systems that drive most of the internet actually work.
- AI safety & alignment — AI Safety Fundamentals Free, structured courses on making powerful AI systems safe.
Practice & compete
Sharpen your instincts against real problems and other people’s solutions.
Read the field
Stay current and build depth with the writing practitioners actually read.
- Lil’Log (Lilian Weng) Deep, clear explainers on diffusion, attention, agents, and more.
- The Batch — DeepLearning.AI A weekly, readable digest of what’s happening in AI.
- MIT 6.S191 — Intro to Deep Learning A polished, free university deep-learning course, updated yearly.
- fast.ai — Practical Deep Learning A famously effective top-down, code-first course.
Production & MLOps
Turn notebooks into systems that run reliably for real users.
Join the community
Learning with others keeps you going — ask questions, share projects, get feedback.
Free courses & certificates
Structured, free courses from the best sources — several give a shareable certificate.
- DeepLearning.AI — short courses Free, focused 1-hour courses on RAG, agents, fine-tuning, and more.
- Google — Machine Learning Crash Course A polished, free intro to ML with interactive exercises.
- Kaggle Learn Bite-size, hands-on micro-courses with a certificate each.
- Hugging Face Learn Free courses on NLP, deep RL, audio, and diffusion.
- Elements of AI A gentle, free university intro for the non-technical side of AI.
Books you can read free
Full textbooks, online and free — for when you want depth, not a video.
- Dive into Deep Learning (d2l.ai) Interactive, code-first deep-learning book used at 500+ universities.
- Mathematics for Machine Learning The exact linear algebra, calculus, and probability ML needs — free PDF.
- Understanding Deep Learning (Prince) A modern, beautifully illustrated deep-learning textbook — free PDF.
- Neural Networks and Deep Learning (Nielsen) The classic, intuitive intro to how neural nets learn.
Watch & listen
Channels, newsletters, and podcasts to keep learning in the background.
- 3Blue1Brown The most beautiful visual explanations of the math behind AI.
- StatQuest with Josh Starmer ML concepts explained clearly, one bite at a time.
- Andrej Karpathy — Zero to Hero Build neural nets and GPTs from scratch on video.
- The Batch (DeepLearning.AI) A calm weekly newsletter on what actually matters in AI.
- Latent Space podcast Deep, practical conversations on building with AI.
Datasets & tools
Where to find data to practice on, and the platforms practitioners actually use.
- Hugging Face Datasets Tens of thousands of ready-to-use datasets, one line to load.
- Google Dataset Search A search engine just for datasets across the whole web.
- UCI Machine Learning Repository The classic home of clean, well-documented ML datasets.
- Google Colab Free notebooks with a GPU — no setup, run everything in the browser.
Contribute to open source
Working in a real codebase with real reviewers is a level-up nothing else gives you.
Turn it into a career
Know the roles, build proof, and get hired.
Keep this page bookmarked — and if you build something great,share it back on GitHub.