AI specialist, application engineer, full-stack: what are you actually hiring for?
The titles have multiplied faster than the clarity. Before you write the job post, it helps to separate the label from the work you actually need done.
Summarize this article with AI
ChatGPT
Claude
PerplexityKey takeaways
- Job titles overlap and mislead; start from the work the product needs, not the label.
- Most 'AI engineer' roles are really application engineering with a model wired in.
- First products usually touch three or four specialties, which is the case for a team over one hire.
Founders ask us, almost weekly, whether they should be hiring an 'AI engineer' or an 'application engineer' or a 'full-stack developer.' Underneath the title confusion is a simpler question they haven't separated out: what does the first version of this thing actually require?
The titles, briefly demystified
- Full-stack engineer. Builds the whole application, front to back. The closest thing to a one-person product team, which is why it's the most common first-hire target.
- Application engineer. Often used for someone who builds on top of existing platforms and APIs rather than designing systems from scratch. Great for assembling, less so for architecting.
- AI engineer / AI specialist. A loaded term. It can mean someone who fine-tunes models (rare and expensive), or, far more often, someone who wires LLM APIs into a product. Most companies need the second and accidentally write a job post for the first.
The labels overlap, and the market uses them loosely. A 'senior full-stack engineer' at one company is an 'application engineer' at another doing identical work.
Start from the work, not the title
Write down what the product has to do in its first six months. Then map it to skills, not job titles:
- A customer-facing web app → front-end plus back-end plus a database. That's full-stack work.
- An internal tool that pulls from your existing systems → integration and API work.
- A chatbot, a voice agent, or document automation → LLM integration, prompt engineering, and the plumbing around it. This is 'AI' in the way most companies mean it, and it's mostly application engineering with a model in the loop.
- It needs to stay up and not leak data → infrastructure and security, which almost no single 'first hire' is genuinely senior in.
Specialties a first product usually needs
Share of early builds that require each area
Now count the specialties. Most first-version products touch three or four. No single hire is genuinely senior across all of them, which is the quiet reason so many first hires underdeliver: you asked one person to be four.
The 'AI engineer' trap specifically
AI talent is the hottest, scarcest, most expensive hire on the market right now, and most companies asking for it need an application engineer who is fluent with modern model APIs, not a researcher. You can pay a fortune for the wrong specialist, or you can find a strong product engineer who ships AI features. We almost always recommend the latter.
Hire for the work the product needs, not the title that's trending. The title is marketing. The work is the job.
If the honest answer is 'we need three or four specialties and we're not sure which matters most yet,' that's the textbook case for a team over a hire. You get the full skill set without betting the role on one person's resume.
