Interview Coach

Realistic practice · Clear feedback · Your control

Your next answer can be stronger.

Speak naturally and think out loud. Your coach follows your reasoning, asks sharper follow-ups, changes the difficulty, and shows you exactly what to improve before the real interview.

Try the hosted alpha Run it locally instead
  • Naturalanswer in your own words
  • Adaptivequestions follow your answer
  • Your keystart keyless or bring your own
Your practice interview 07:42

Principal engineer · Systems design

“Why choose an event stream here instead of a simpler queue?”
Your next question Advanced ↑
Coach follow-up

Pause there. What failed first under load?

Practice how you actually speak.

Find where your answer loses force.

Leave knowing what to try next.

Feedback you can use

Know what landed. Know what to improve next.

Your report connects every assessment to something you actually said. You get specific evidence, practical next steps, and no made-up pronunciation score when the session has no usable audio evidence.

01

Confidence

See where you sound decisive—and where your answer needs a clearer choice, assumption, or outcome.

02

Communication

Learn whether your answer is easy to follow, relevant, concise, and right for the audience.

03

Technical depth

Strengthen the trade-offs, constraints, failure modes, and proof behind your technical decisions.

04

Pronunciation

Get this signal only when supported acoustic evidence is available. Otherwise, it stays unassessed.

Control before convenience

Useful coaching should not require blind trust.

Start in deterministic demo mode without an API key. If you choose model-backed evaluation, bring your own key. The scoring rules are inspectable, your sessions can be exported or deleted, and the application never treats a coaching score as a hiring decision.

See how your data moves
Browser UI Next.js API LangGraph.js
Local evaluator OpenAI adapter PostgreSQL
Your provider key stays out of interview state. Keep it in the browser tab or opt into encrypted per-user storage when self-hosting.

Honest about what works

Start with a real practice loop—not a product promise.

What you can use today

  • Practice a five-turn adaptive interview
  • Try the complete flow without an API key
  • Choose optional OpenAI-backed evaluation
  • See evidence behind every dimension score
  • Review a recruiter-style coaching report
  • Pause now and resume your session later
  • Return with an optional pseudonymous recovery account
  • Export or permanently delete your interview
  • Run the full application with one command
  • Practice against six versioned role rubrics
  • Use text mode as the reliable default
  • Explore a consent-gated WebRTC voice beta

What we are proving next

  • Voice reliability across real browsers and devices
  • Score calibration with qualified human reviewers
  • Verified identity and team workspaces
  • More deeply calibrated role packs
  • Bias and score-stability evidence
  • Independent WCAG 2.2 AA review
  • Reliability evidence for a hosted service

Start on your machine

Your practice room is one command away.

Run the complete application on your own machine. Demo mode needs no API key; add your own OpenAI key only when you want model-backed evaluation or the experimental voice experience.

terminal · project root 
$ docker compose up --build --detach --wait
✓

Your local coach is ready
http://localhost:3000

Open source means inspectable

Know what you are trusting before you practice.

Read how the coach works, what each score means, where your data goes, and which capabilities still need stronger evidence.

01 See what the product is designed to do Candidate journeys, coaching requirements, scoring, safety, and success measures. 02 Understand how your data moves System boundaries, key handling, interview state, voice, and testing. 03 Follow what is coming next Alpha, durable sessions, real-time voice, and public-v1 exit criteria. 04 Understand every coaching score Score evidence, interruption policy, public fixtures, and known calibration gaps. 05 Review the license status The current license status and options still requiring a maintainer choice. 06 Check the readiness evidence Verified evidence, missing controls, external validation, and release gates. 07 Review the security model Assets, trust boundaries, attacker stories, invariants, and severity calibration.

Make the real interview feel familiar

Practice until your reasoning sounds as strong as it is.

Run Interview Coach on your machine, rehearse the role you want, and turn every attempt into a clearer story. If something is missing, help shape it in the open.

Get the open-source project Request a feature

Interview Coach provides practice and decision-support signals. It is not a hiring decision, a clinical assessment, or a substitute for qualified human review.