Open beta · free
Learn AI without starting over
AI-grafen finds what you already know and what you are missing, and builds the shortest path to where you want to go — with labs that test your code and proof that you can do it.
41 goals · 509 knowledge nodes from the basics to the research frontier · 29 labs and projects · Swedish and English
Choose your starting point
I'm a developer →
Skip what you know. RAG, agents, evals and fine-tuning — with labs where tests decide whether the code holds up.
I already work with AI →
Find the gaps and show what you can do: evals, agents, operations and the research tracks at level F–G.
Product or management →
Understand what models can and cannot do, without code, at the level you need to make decisions.
I'm starting from zero →
Python, statistics and neural networks in an order that builds on itself. The diagnostic finds where you stand.
How it works
1. Pick a goal
«Understand generative AI» or «train a neural network», say. The goal decides which knowledge is needed.
2. The diagnostic
Questions that search along the prerequisite chain — you skip what you know and start at your first real gap.
3. The learning path
Nodes in the right order. Every node has intuition, formulas and code. The path is recomputed as you learn.
4. Mastery
Nothing counts as known because you read it. Exercises, explanations in your own words and labs with hidden tests decide.
The diagnostic finds your first gap
The questions follow the prerequisite chain backwards from your goal. The result shows what you already know and where your path starts.
A learning path that recalculates
The steps come in the order the knowledge builds on itself. Master a node and you move on; get stuck and you get a detour through what is missing.
A tutor that does not give away the answer
The AI tutor asks questions back and gives hints in steps, based on the node's content and sources.
Labs with real tests
Write the code yourself in the browser or in a sandbox on the server. The tests decide whether it holds — not whether it looks right.
What professionals learn here
Build a RAG system you can trust
EChunking, retrieval evaluation, citation and grounding — the whole chain from document to answer with traceable sources.
Build an agent you can trust
FArchitecture, planning, observability, safety and evaluation — and the coding agent that gets objective feedback from tests.
Evals in practice
FLLM as judge, regression tests, statistical significance and how to value negative results — measurement that holds between releases.
Fine-tune a model with LoRA
FUnderstand why low-rank adapters work, choose target modules and rank, build a dataset, and evaluate the result with evals.
Show all 41 goals
- BAI for beginners
- DUnderstand how generative AI works
- DTrain your first neural network
- EBuild a transformer from scratch
- FBuild a memory system for an agent
- GReproduce a paper
- EBuild an NLP system end to end
- GFrontier Lab — an independent research project
- CComputers that see and hear
- EImage classification with convolutional networks
- ETrain an agent with reward
- EBuild an AI service that survives production
- FRun models more cheaply: quantisation
- EThe mathematics behind the models
- CFrom blocks to Python
- DThe developer's toolbox
- ETraining neural networks for real
- DClassical machine learning in practice
- EStatistics for experiments
- ELanguage models in practice
- CAI, ethics and society
- DData: collect, clean, document
- DFoundations of computer science
- ESeeing and hearing with AI
- EAn AI service in operation
- FFine-tune and run your own models
- FAI safety in practice
- FGenerative models in depth
- FAI in production
- ESystems knowledge for AI engineers
- EClassical ML for real
- FInterpreting a language model
- EResponsible AI in practice
- FBuild a voice interface
- GMultimodal systems
- FDeep reinforcement learning
- DThe language of mathematics in AI texts
Frequently asked questions
Why not just ask ChatGPT?
A chatbot answers what you ask well. It doesn't know what you already know, what you are missing or whether you actually learned what it explained. AI-grafen keeps track of that, picks the next step from it and lets exercises and tests decide.
What does the certificate mean?
It lists the knowledge you have shown you master, the kinds of evidence behind it and the labs you passed. It is signed and can be verified. It is documented evidence — not a formal qualification.
Can my team use AI-grafen?
Yes. Everyone can create a free account. Contact us to try a shared skills map and goals for your team. We are not charging at the moment.
Who is it for?
Adults who want to understand or work with AI — from people who have never programmed to people who want to reproduce research papers. The diagnostic decides where you start.
Do I need to know how to program?
No. The early goals need no code, and there are paths that teach you Python from scratch. Code appears when it is needed.
What does it cost?
Nothing during the beta. We will tell you well in advance before anything changes.
What happens to my data?
It is kept on our own server, used only to provide the service and understand whether it works, and you can export or delete everything whenever you want.
How do I get started?
Create a free account and choose a goal. No invitation code or payment needed. You can also try the diagnostic right away without an account.
See where you stand — in ten minutes
Pick a goal and take the diagnostic without creating an account. You see right away what you already know and where your path would start.