DevOps Interview AI: One Idea, Three Platforms

#android #webdev #ai #kotlin

DevOps Interview AI runs realistic mock DevOps job interviews entirely on-device — no backend, no account, no anxiety of facing a real human interviewer. The project grows in parallel across three platforms — Android, Web, and iOS/macOS — and the latest round of changes brought Android and Web to a shared visual and functional standard.

Here's what changed.

📱 Android: voice that no longer cuts you off, and design parity with the web app

The most painful recent bug: voice input would cut the candidate's answer off mid-sentence. The speech recognizer on some devices (a Samsung tablet, specifically) ignored the app's silence-timeout hints and decided "answer done" far too early.

The fix: instead of trusting the recognizer to decide when you're done, SpeechToTextManager now restarts the recognition session itself every time the device ends it prematurely, accumulating what was actually said. The answer is only sent once real silence follows actual speech. A live "Listening…" caption shows the candidate they're genuinely being heard.

The second focus was design parity with the web app: a Home button and an EN/UA language toggle now sit on both the interviewer-selection screen and the call screen, switching language instantly with no app restart.

Third: a new avatar for HR persona Olivia — "Lil Avatar" by lilchen, CC BY 4.0 — replacing the old rig's baked-in background.

Release: v1.10.0

🌐 Web: a full 5-stage hiring pipeline instead of a single chat

The web app, meanwhile, grew from a simple chat into a full DevOps hiring flow. The home screen now surfaces every mode up front instead of dropping you straight into a chat:

5 sequential interview stages — Recruiter Emma → Senior DevOps Marcus → CTO David → Project Manager Olivia → a final Offer, each stage unlocking only after the previous one is complete

Practice & self-prep — a question bank per persona plus a self-test quiz

Resume review — an automated improvement checklist, entirely client-side, nothing sent to a server

Job search resources — a curated list of resources and tips
The same voice-input fix as Android — the recognizer no longer cuts an answer off mid-thought

What ties both platforms together

Both apps now share one design language (glassmorphism cards, a Home + EN/UA top bar) and the same approach to voice reliability — don't blindly trust the recognizer's idea of "end of thought"; verify real silence after real speech instead. A small detail that directly decides whether a candidate feels like the app just interrupted them mid-interview.