Related Posts
How is IBM for Full Stack Java developer role?
Additional Posts in Ask A Recruiter - Tech
Do cover letters matter at all anymore?
Any recruiter in search of a Scrum Master 🙂
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.



Mentor
This has been stressing me out too. I’m not sure if I need to dive deep into LLM architecture or just nod along in interviews. Would love clarity from recruiters on what “AI experience” actually means in these listings—are we talking prompt engineering or TensorFlow from scratch?
For most backend/fullstack roles, it’s usually HR or hiring manager buzzwords. You rarely need to know the deep mechanics of LLMs. They’re mostly looking for how you can leverage AI tools like Copilot or ChatGPT to improve productivity, not build them from scratch.
Unless you are an AI developer/data scientist (the ones who build models) you only need to know the high level.
The "gotcha" I ask to figure out how much you know about using AI, is asking how do you prevent yourself from falling trap to AI hallucinations?
If the average AI user is using it for information, and that information WILL include some hallucinations, how do you protect yourself from it? If you have no answer then you'd auto fail my gotcha. If you have ANY method then you are at least aware of the main most common shortcoming to using LLM based AI.
I only interview developers, and I've heard answers ranging from:
- check with other models for validity
- ask the model to provide a rating for accuracy
- verify utilizing non-AI methods (testing)
- don't ask for anything you can't verify another way
heck even "manually test everything", is valid. It isn't optimal, but at least you are aware.
The bad answers I've heard are along the lines of:
- tell it its wrong (this doesn't do anything, and assumes you know its wrong)
- ask a different model (this avoids the problem)