Practice interviews with a live AI video interviewer, then get a scored report on how you did. For anyone prepping for a job interview and wanting reps before the real thing. Built on the DeepSpace SDK.
Live app: https://mock-interview.app.space
- Runs a real-time mock interview against a talking video avatar that asks questions and follows up
- Covers different interview styles (behavioral, coding, and more)
- After you hang up, turns the transcript into a scored report that grades each answer
- Keeps a history of your past interviews and reports
The face-to-face interview is driven by the tavus integration through the
DeepSpace integrations proxy — the app creates a persona and a conversation, then
polls it for the finished transcript. Scoring runs as a durable JobRoom job
(score-interview): once the call ends it waits for the transcript, sends it to
the anthropic integration to grade each answer, and writes the reports row,
flipping the interview to "scored." The report page subscribes with the useJobs
hook to show live progress, and interviews and reports are stored in RecordRooms
with per-collection RBAC.
Deploy your own copy in three commands:
npm install
npx deepspace login # one-time, opens a browser tab
npx deepspace deploy # -> <name>.app.spaceAuth, the database, real-time sync, and hosting all come from DeepSpace, so
there is nothing else to configure. Your subdomain is the name field in
wrangler.toml; change it for your own deployment.
Or build something new: apps like this are made by handing a prompt to a
coding agent — start at deep.space/get-started,
or scaffold from scratch: npm create deepspace@latest my-app.
Dialogue was built end-to-end by an AI agent on the DeepSpace SDK. DeepSpace is laying the foundation for rebuilding the Internet in an AI-native way — deep.space · docs.