Open source & self-hostable★ Star us on GitHub →

Open-source AI interviewer

Practice the interview out loud.
Then pass the real one.

DeepInterview reads your CV and the job description, researches the company, and runs a real voice mock interview — then shows you exactly what to fix. English-first, available in 10+ languages.

No credit card Self-host with your own keys Apache 2.0

Prepare for interviews at companies like

GoogleAmazonStripeGrabAtlassianDatadog

How it works

Prepare, perform, improve.

The heavy thinking happens before the call, so the conversation itself stays fast and natural.

01 — PREP

Drop your CV & the job post

We extract your experience, the role's real requirements, and the gaps between them — then research how that company actually interviews.

02 — INTERVIEW

Talk to your interviewer

An adaptive voice interview with follow-ups, interruptions, and behavioral, technical and coding rounds — in the language you choose.

03 — IMPROVE

Score, then study

A per-skill scorecard with model answers, and a coach that teaches your weak areas before you go again.

The product

A real conversation, not a quiz.

Real-time voice

Speak naturally. Get interrupted. Recover.

Sub-second responses, barge-in, and a full transcript of every answer. Speech recognition is chosen so accented English is understood, not penalized.

  • Behavioral, technical & coding rounds
  • Live captions & saved transcript
  • Choose an interviewer persona
LANGUAGES
EnglishTiếng ViệtEspañolहिन्दीBahasaPortuguêsFilipinoFrançais+ more
English-first, with first-class interviews in your own language — drill the same role in both.

Personalized

Questions built from your CV, the JD & the company.

No generic question bank. A planner agent researches the role and the employer, finds where your experience is thin, and shapes a difficulty curve around it.

  • CV ↔ JD gap analysis
  • Company-specific question patterns
  • Adaptive follow-ups
QUESTION PLAN · SENIOR BACKEND
Distributed systemsprobe
Kafka & event streaminggap
System design — paymentscore
Behavioral — ownershipcore
SQL window functionswarm-up

The loop

Practice, feedback, study — repeat.

Every interview is scored per competency. Your weak areas flow straight into a document-grounded study coach with flashcards and a voice tutor — then back into the next interview.

  • Grounded answers with citations
  • Spaced-repetition flashcards
  • Mastery tracking across sessions
1CV + JD → gap analysis
2Study plan for weak areas
3Voice mock interview
4Scored feedback
back to step 2, sharper

Open source

Run it yourself in one command.

The full voice pipeline, multi-agent brain and study coach are open under Apache 2.0. Bring your own model keys, add a language pack, or contribute back.

# clone & run the full stack
$ git clone github.com/ngoanpv/DeepInterview
$ cd DeepInterview
$ docker compose up
✓ web → http://localhost:3000
✓ agent → connected (LiveKit)
✓ ready → start an interview

FAQ

Good questions.

Which languages are supported?

English is the default, with first-class support for 10+ languages including Vietnamese, Spanish, Hindi, Bahasa, Portuguese and more. Each language is a pluggable pack, so coverage keeps growing — and you can interview for the same role in two languages.

Is it really free and open source?

Yes. The core is open source under Apache 2.0 — self-host it with your own model keys and own your data end to end. Add a language pack or new interviewer persona and contribute it back.

What are the avatars?

Original, IP-safe video personas — an anime mentor, a hero-mode coach, a professional recruiter — that react to the conversation. They're pre-rendered, so they stay smooth on any device.

Ready when you are

Your next interview starts here.

Upload a CV, paste a job post, and talk it through — in minutes.