Let's start by reviewing our Q2 outcomes. We hit 87% of our OKRs — slightly below target but acceptable given the resource constraints we faced in the last sprint.
Agreed. Feature velocity was lower than expected, but quality metrics improved significantly. User retention went up 12% quarter-over-quarter, which is a strong signal.
For Q3, I think we should prioritize the real-time collaboration module. It's been requested by 68% of enterprise customers in the last survey and directly ties to our retention goal.
Good point. We should also revisit the performance optimization work that's been deferred. The mobile team is seeing latency issues directly affecting conversion rates.
I can take ownership of the collaboration module spec. We could have a draft ready by end of next week and open it for team review before locking scope.
I'll handle the performance audit. Let me coordinate with the infrastructure team and come back Thursday with timeline estimates and resource asks.
Meetings Are Broken — and Nobody Knows While It Is Happening
Remote and hybrid work has made video conferencing central to global collaboration. Yet moderators have no real-time visibility into participant engagement, focus levels, or agenda drift. Existing tools capture audio only, and insights arrive only after the meeting has already ended.
A Standalone Desktop Application That Sees, Hears, and Understands
MeetAI is a standalone Electron.js desktop client that operates alongside any video conferencing platform — without requiring API integrations, bot invites, or vendor agreements. It captures screen frames and system audio directly at the OS level, routes data through three parallel AI pipelines running on our own infrastructure, and delivers live coaching alerts to the moderator in real time.
Raw audio and video streams are processed server-side and immediately discarded after feature extraction. No media data is stored or retained at any point.
Three Parallel AI Pipelines
What Existing Tools Get Wrong
Current solutions are reactive, audio-only, and platform-bound. MeetAI is the only solution combining multimodal sensing, real-time intervention, and privacy-by-design in a single deployable desktop client.
| Feature | Bot-Based Tools | Native Platform Tools | MeetAI |
|---|---|---|---|
| Data Analysis | Audio and Text Only | Audio and Text Only | Audio + Visual + NLP |
| Intervention Timing | Post-Meeting Only | Post-Meeting Only | Real-Time Alerts |
| Platform Dependency | Bot Integration Required | Vendor Lock-In | Independent Desktop Capture |
| Attention Tracking | None | None | Head Pose + MAR |
| Data Privacy | Cloud Storage | Cloud Storage | No Media Retention |
Three-Layer System Design
MeetAI separates concerns across three fully isolated layers, enabling independent scaling and hot-swappable ML models without modifying the core orchestrator.
-
Frontend — Electron.js + React + TypeScriptOS-level media capture, real-time dashboard rendering, live alert display.
-
Core Backend — Node.js / Express.jsSession lifecycle management, sliding-window data fusion, rule engine, SSE alert delivery.
-
AI Workers — Python 3.xIsolated Vision, Speech, and NLP processes on our own infrastructure via Redis Pub/Sub. Zero shared memory between pipelines.
What MeetAI Demonstrated
Meet the People Behind MeetAI
We are senior Software and Computer Engineering students at Işık University. Reach out — we'd love to hear from you.
