An earlier shape of this project rendered a digital stand-in that spoke for you. It was built, and then deliberately dropped: an undisclosed AI stand-in deceives the employer, and the market already rejects it — recruiters disqualify candidates who show up as renders, and interview platforms ship gaze and voice-modulation detection.
What remains is a copilot. It never speaks, never clones a voice and never invents an achievement — it can only surface things you already wrote down about work you already did. The distinction is the whole design.
Highlights
- Does not clone your voice, render an avatar, or speak on your behalf
- Hybrid retrieval over a private knowledge base you control
- Deepgram streaming speech-to-text into Groq Llama-3.3-70b
- Three or four bullets per question, sized to be read at a glance
What you see on the second screen
interviewer asks → "How did you handle cluster backups?"
▸ etcd snapshots to object storage, nightly
▸ restore rehearsed quarterly — ~8 min end to end
▸ first rehearsal caught uploads silently failingIllustrative — the MVP is still being validated, so there are no screenshots of a real session yet.
Need something like this built?
Tell us what the app has to do and where it has to run. You'll get a scope, a timeline and a price.
Start a project