oVice
Led the Virtual Meeting Room redesign, plus meeting logs, co-editing and performance work for 4,000+ companies.
- Year:
- 2024
- Role:
- Full-Stack Software Engineer
- React
- TypeScript
- Jotai
- SWR
- TanStack Query
- Elixir Channels
- Phoenix
- Ecto
- PostgreSQL
- Yjs (CRDT)
- ExUnit
- Vitest
- Playwright
Full-stack feature work on oVice's spatial, avatar-based virtual workspace, used by 4,000+ companies — OpenSpace for walk-up-and-talk collaboration and private, bookable Meeting Rooms for focused discussion. Owned Meeting Logs on a Yjs/BlockNote real-time co-editor with secure playback for 100,000+ users, shipped meeting-room hosts across React and Elixir/Phoenix, led the Meeting Room revamp, and tuned frontend performance across the platform.
Problem
oVice's virtual workspace serves 4,000+ companies across OpenSpace — the shared space where avatars move freely and colleagues can walk up and talk without scheduling a call — and private, bookable Meeting Rooms for focused discussion. The core Meeting Room's design had plateaued in daily adoption, admins had no record of who joined, left or chatted in a session, users had no way to co-edit or securely revisit recordings, and accumulated bugs and render-heavy UI were slowing everyday interactions down.
Approach
Owned Meeting Logs end to end: filters, drafts, agendas, import and copy on a Yjs/BlockNote real-time co-editor (copies read the live document across the Phoenix cluster), with Phoenix endpoints, Ecto migrations on PostgreSQL and ExUnit tests across 22 test files, plus secure playback for recorded meetings. Shipped meeting-room hosts full stack — REST endpoints, automatic host assignment and late-joiner sync over Phoenix Channels, and the UI — and led the redesign of the core Virtual Meeting Room. Optimized the frontend with optimistic UI patterns and targeted React/Jotai render tuning.
Outcome
The Meeting Room revamp drove a 30% increase in daily VMR adoption, 100,000+ active users get a collaborative meeting record and secure access to recordings, and the performance work cut interaction latency by 25%.