ZVSquared: The AI interviewer that won't accept memorized answers

Hey Product Hunt — I’m Aleksandr, founder of Zvsquared.

I started building Zvsquared after failing a technical interview on a question I thought I understood: logistic regression.

The interviewer interrupted me and asked:

“Why the log? Why specifically that function?”

I knew the standard explanation. But once they kept pushing, I realized I couldn’t defend the reasoning underneath it.

That stuck with me.

AI and ML interviews are full of this. It’s easy to memorize what cross-entropy is, explain RAG at a high level, or repeat the usual bias-variance answer.

A strong interviewer doesn’t stop there.

They ask why. Then what breaks. Then what you would change in production. Then why you chose that trade-off instead of another.

Zvsquared is the interviewer I wish I had.

It runs timed, text-based mock interviews for AI and ML engineers. Instead of accepting a plausible first answer, the examiner keeps probing your reasoning, identifies exactly where your understanding breaks, and turns those gaps into a practice plan.

At launch, you can practice topics including:

— Machine learning foundations
— LLMs and AI engineering
— RAG and retrieval
— Probability and statistics
— Linear algebra
— Technical reasoning and system trade-offs
— Timed mock interviews with adaptive follow-up questions
— Saved transcripts, missed criteria, feedback, and progress over time

This is deliberately not a chatbot that gives you the answer as soon as you struggle.

The goal is to expose the gaps before the real interviewer does.

I’d especially value feedback from people preparing for AI Engineer and ML Engineer interviews:

What would make the interviewer feel more realistic, more difficult, or more useful?