roadmap.sh: AI Engineer

What it is

The AI Engineer roadmap is one of the community-maintained roadmaps on roadmap.sh, a project with about 368k GitHub stars and more than 3.2 million registered users as of September 2026. It is a clickable map: each node opens a short explanation and a list of free and paid resources for that topic. Signed-in users can mark topics as done and track progress, and the page links project ideas and an AI tutor.

The distinction it makes well

The roadmap defines an AI engineer as someone who uses pre-trained models and existing AI tools to build useful products, without building models from scratch. That is different from ML engineers and researchers, who train models and develop theory.

This distinction matters. The large majority of jobs advertised as AI roles are the first kind, and many people spend months on the second path (heavy mathematics, training from scratch) because nobody drew the line for them.

What it covers

The topics follow the work of building AI products:

• Foundations: what language models are, common terminology, and choosing between hosted and open models.
• Using model APIs and prompt engineering.
• Embeddings and vector databases.
• Retrieval-augmented generation (RAG) for answering from your own data.
• Agents and tool use, including protocols for connecting models to tools.
• Multimodal AI: images, audio and speech.
• Evaluation, safety, observability and cost, the topics that separate demos from production.

How to use it

Use it as an audit and a job description decoder, not a syllabus to complete top to bottom:

1. Tick off what you already know.
2. Collect five job postings for roles you want and highlight skills they mention.
3. Compare the gaps with the roadmap and order your study by what employers ask for.
4. Build one project per major section, such as a chatbot using an API, a RAG app over your notes, an agent with two tools, then add evaluation to one of them.

The linked resources vary in quality; pick one per topic rather than working through all of them.

Before you start

You need solid programming in Python or TypeScript and comfort with APIs, JSON and basic web development. If you are still learning to program, the backend or frontend roadmaps are better first steps.

Where it falls short

It is a map, not a course, so it tells you what to learn but not in depth how. The field changes quickly, and some nodes lag behind the newest tools. It also cannot tell you which topics a specific employer weights most; job postings do that.

Common questions

Do AI engineers need a degree in machine learning? The roadmap's FAQ addresses this; in practice, strong software engineering and shipped AI projects matter most for AI engineering roles.

How long does it take? For an experienced developer, a few months of focused building covers the core. Beginners should first become solid developers.

Is it free? Yes.

Best for: engineers planning a move into AI product work, and anyone decoding what an AI job posting expects.