Back to index
OtherSelf paced, with the AI Engineer path running roughly 30 to 40 hours·Free tier available, Pro subscription around 30 dollars per month or discounted annually

Scrimba (Interactive Coding and AI Engineer Path)

4.1

Scrimba's interactive screencast format is the most underrated idea in online coding education, because you learn by editing real code inside the lesson rather than watching a video in one window and typing in another. The AI Engineer path is a solid, practical bridge for web developers moving into LLM work, though it is lighter on machine learning fundamentals than the name might suggest.

What We Liked

  • The interactive scrim format genuinely aids retention, since you can pause and hack on the instructor's code at any moment
  • Excellent for hands on, build first learners who lose focus during long passive lecture videos
  • The AI Engineer path is refreshingly practical, focused on calling APIs, embeddings, RAG, and shipping real apps
  • A friendly, active Discord community and a price that is reasonable for the amount of interactive content

What Could Be Better

  • It is application focused, so it teaches you to use AI models far more than it teaches how they work under the hood
  • Deep theory, maths, and model training are largely out of scope, which suits builders but not aspiring researchers
  • The platform's heritage is frontend web development, and that DNA still shows in the framing of everything
  • The interactive format works less well for very heavy data or notebook style workflows than for web code

Detailed review

Scrimba is one of the few online learning platforms that has an actual technical innovation at its heart rather than just a content library, and having spent time inside its lessons I remain a fan of the core idea even where I have reservations about scope. The signature format, which they call a scrim, is an interactive screencast where the instructor is talking and coding, but at any point you can hit pause, click straight into their editor, change the code, run it, and experiment, before resuming. That sounds like a small thing and it absolutely is not, because it collapses the awkward gap that plagues normal video courses where you watch in one window and try to replicate in another and constantly fall behind. For a hands on learner who thinks by tinkering, this is close to ideal, and it is the single biggest reason I would point someone toward Scrimba over a conventional video course.

The newer AI Engineer path is a sensible evolution of the platform, and it is important to understand what it actually is, because it is an application engineer's route into AI rather than a machine learning theory course. It concentrates on the things a working developer genuinely needs to ship AI features, such as calling model APIs, working with embeddings, building retrieval augmented generation over your own data, handling prompts sensibly, and wiring all of that into a real deployable app. That practical framing is exactly right for its target audience and I respect that it does not pretend to be something loftier. The flip side, and the thing I would want a prospective student to be clear eyed about, is that it will teach you to use models far more than it teaches you how they work, so the maths, the internals of a transformer, and anything resembling training your own model from scratch are simply not the point here.

Scrimba's roots are in frontend web development, and that heritage still colours the whole experience, which is a strength if you are coming from that world and a slight mismatch if you are arriving from a data science or research background expecting notebooks and statistics. My verdict is that Scrimba is a genuinely delightful way to learn for the right person, the web developer or coding beginner who wants to build AI powered applications and learns best by doing, and I would happily send that person here, while gently steering the aspiring researcher or the person who needs deep fundamentals toward something like Fast.ai or Andrew Ng instead.

[ final ]

The verdict.

A strong recommendation for a specific journey, namely a web or frontend developer who wants to start shipping AI powered features without wading through linear algebra first. If your goal is to understand transformers from the ground up or train your own models, this is the wrong tool, but for practical AI application building the interactive format is a real joy to learn with.