Right now you hear a lot about “AI Engineering”, companies and recruiters are searching for “AI engineers”. But there seems to be a big confusion about what AI engineering actually is.

There are two fundamentally different meanings behind this word:

• Engineers who are “AI first”, aka actively using AI programming tools such as Cursor, Claude Code, etc.
• Engineers who are building AI systems, or to be more precise, integrating LLMs into existing software applications.

So let’s break this down in detail:

Engineers who are AI first

This is the first class of people who are often being called “AI engineers”, which in my opinion is pretty much misleading. If you are reading through platforms such as X, HackerNews, etc., you see that more technical people are actually calling this group “vibe coders”. I also don’t like the term “vibe coder”, but it’s not misleading at least.

In short, vibe coders favor using either an AI-only approach towards software engineering/programming or they offload the majority of their work to AI. In the future, I will write a specific blog post about vibe coding.

In hiring right now (November 2025), there are recruiters who actually want to hire vibe coders (positions specifically describing an AI-first approach towards programming). But oftentimes, these positions are labeled as “Hiring AI engineer”.

Building AI Systems

The other group of “AI Engineers” are software developers who are integrating LLMs into existing applications or building completely new applications around LLMs. These can be either RAG-based chatbots, workflow automation, agents, or other kinds of LLM-centric apps.

Chip Huyen has described it actually pretty well in her book “AI Engineering”:

“AIE focuses on building applications on top of foundation models, which involves more prompt engineering, context construction, and parameter-efficient finetuning.”

— Chip Huyen, 2025

This is by far the best summary of WHAT AI Engineering actually is. It should be noted though that there are two distinctions that should be made when talking about this group:

• “AI Engineering” does not mean machine learning, MLOps, or ML Engineering. It’s very important to distinguish that.
• Although AI engineering sets a strong focus on LLMs, the majority of AI-centric software systems are still 80% (if not more) about “classic” software engineering principles: data processing, architecture, auth, data storage, monitoring, etc.

The AI aspect here typically revolves around: prompting, making sure the output from the LLM matches the expected output, choosing the right models, certain concepts such as “LLM as a judge”, etc. There are definitely many challenges around AI Engineering (with fine-tuning being probably the hardest one, that also comes closest to machine learning).

I hope that in the future, for hiring, recruiters will start to distinguish between “AI Engineering aka building AI-centric software” and preferring developers who have an AI-first approach.