Interviews that interrupt.
Reading answers doesn't teach you to speak under pressure. A real-time voice interview asks about your resume, follows up on vague answers, and scores how you said it.
3 rounds · Resume-aware · 5-minute rounds free · No card
Never asked of the model, so the numbers can't be hallucinated.
Walk me through the caching layer on the delivery app you listed.
Structure: opened with the fix, never stated the problem.
Red flag: couldn't name a tradeoff for your own decision.
Three rounds, one interviewer
HR / Behavioural
Motivation, conflict, teamwork. Scored for structure, not vibes.
Resume Grill
Reads your uploaded PDF and asks about your real projects — why that choice, what broke, what you'd redo.
CS Fundamentals
Indexing, concurrency, complexity, API design — spoken aloud, no code editor.
It separates two things deliberately
How you said it is arithmetic. Whether the answer was any good is judgement. Mixing the two is how feedback becomes noise.
Filler-word counts, words per minute, speaking time. Never asked of the model, so the numbers can't be hallucinated. Pace is measured against time actually spent talking, so a nervous pause doesn't read as slow speech.
- 17
- Fillers
- 168
- WPM
- 9:12
- Talk time
A score out of 10, per-answer "what you said → what a strong answer sounds like", structure notes, and red flags.
"We added Redis because the database was slow and it got a lot faster."
"Reads outnumbered writes 40 to 1 and p95 was 900ms. We cached the payload for 90 seconds — p95 went to 60ms, and we accepted a 90-second staleness window because menus change daily."
Nine of twelve answers opened with the solution. No numbers volunteered until asked.
Couldn't name a tradeoff for a decision you made yourself. Credited "the team" for your own work.
A vague answer doesn't get a polite nod
It follows up when an answer stays fuzzy, and it cuts in when one keeps going after the point is made — the way a tired interviewer on their fourth call of the day would.
How it worksWalk me through the caching layer on the delivery app you listed.
So, um, basically we had a lot of read traffic and, like, the database was kind of struggling, so we…
Hold on. What was the read-to-write ratio?
Roughly 40 to 1. We put Redis in front and cached the menu payload for 90 seconds.
Why 90 seconds and not 10?
- Fillers
- 3
- WPM
- 184
- Talk time
- 0:44
As easy to start as a phone call
Headphones in, resume already uploaded, ten minutes on the walk home. Every scorecard stays in your history, so the next round can be compared to the last one.
Get early accessPractise the part you can't practise by reading
LetsMock is in closed beta on Android. Leave your email and you'll get an invite before the Play Store listing goes live.
Android first · No card required in beta