Most "AI engineer roadmap" posts are a wall of buzzwords with no actual curriculum behind them. This one links directly to four free, video-based learning paths — in the order we'd actually recommend working through them — each with real projects, a completion exam, and a shareable certificate at the end.

Expect 8-12 months of consistent effort if you're starting from general programming knowledge, less if you already write code daily. You do not need a PhD, and a portfolio of real, deployed projects will get you further than a certificate alone — which is exactly why every path below is built around building something, not just watching video.

Step 1 — AI-Assisted Development (Vibe Coding)

Start here even if you already know how to code. Every AI engineering role today assumes fluency with AI coding agents — Cursor, Claude Code, GitHub Copilot — plus the judgment to review what they produce. Skipping this step is the single most common reason people build things that don't survive contact with real users.

6 modules · 6 videos · free certificate — Start the Vibe Coding path →

Step 2 — AI Agents Engineering

Once you can work with AI coding tools, move to building with AI agents yourself: how agents plan, use tools, maintain memory, and chain steps together to complete a task without a human driving every action.

8 modules · 12 videos — Start AI Agents Engineering →

Step 3 — Generative AI Engineer, Senior Prep

The deepest path in the set: the theory and hands-on skills behind the LLM applications agents are built on top of — embeddings, retrieval-augmented generation (RAG), fine-tuning, and evaluation.

10 modules · 18 videos — Start Generative AI Senior Prep →

Step 4 — MLOps & LLMOps Engineering

The part most roadmaps skip entirely: how you actually run AI systems in production — deployment, monitoring, versioning models and prompts, and keeping a system reliable once real users depend on it.

6 modules · 7 videos — Start MLOps & LLMOps Engineering →

What to build along the way

  • A small tool you build almost entirely with an AI coding agent, reviewed and shipped by you (proves Step 1).
  • An agent that completes a real multi-step task — booking, research, or data lookup — end to end (proves Step 2).
  • A RAG-based Q&A tool over a real document set you care about (proves Step 3).
  • That same tool, deployed with monitoring and a documented rollback plan (proves Step 4).

Four projects, four certificates, and a portfolio that says a lot more than "completed an AI course." That's the actual roadmap — everything else is just picking which free path to start with today.