What it does:
Live video + a shared code editor both people type in at the same time. The editor sync is a hand-rolled CRDT over a WebRTC data channel, so no server round-trips and no edit conflicts.
Run code in the browser — JavaScript, Python, Dart, C++ — output shown to both sides instantly (via Judge0).
Activity monitoring — the interviewer sees when a candidate switches tabs, opens an overlay, or leaves the app, with a live event log. After 3 flags they can end the interview. Not trying to be Big Brother — just making cheating visible instead of guessed.
Built-in chat, one-tap invite links, and a per-candidate rating + notes history tagged by role.
Stack: Flutter (web + mobile), WebRTC for P2P media/data, Firebase for auth + signaling, Judge0 for execution. Built solo over a lot of late nights.
Honest rough edge: it's fully peer-to-peer, which is great for latency/cost/privacy, but strict corporate/college Wi-Fi firewalls can block P2P — so for now it works most reliably on mobile data / home Wi-Fi. Proper relay (TURN) support is my next big task.
👉 Try it: https://vyntra.space — open it, host a room, send the code to a second tab or a friend, and run a mock interview. Takes about a minute.
What I'd love feedback on: whether the anti-cheat monitoring feels fair or creepy, and what would make you actually switch from your current interview setup.
