It is easy to validate a solution accidentally.
Show someone a polished interview tool and ask whether they would use it, and you are likely to receive polite agreement.
That does not tell us very much.
The first research around Mokchi needs to go further upstream.
How do people currently prepare?
What do they find difficult?
What questions repeatedly cause problems?
Do people practice aloud?
Where do they get feedback?
Do existing mock interview tools feel useful?
What do people do after an interview they believe went poorly?
The strongest evidence will not be feature requests.
It will be behavior.
Someone who spent hours preparing with a friend because they could not judge their own answers is more interesting than someone who says they would probably use an AI coach.
The initial research goal is therefore simple:
Understand the problem before trying to prove the product.