AI Overlay for SUMADI Proctoring on Mac
SUMADI is the AI proctoring platform founded by Raul Rivera in 2019 and now part of the Open LMS / Laureate Education ecosystem. It dominates online proctoring across Latin America and the Spanish-speaking education market, with deployments at large online universities and FUNIBER-affiliated programs.
SUMADI runs as a cloud-hosted AI service on AWS (it landed on AWS Marketplace as a public listing) with two client surfaces on Mac: a browser-integrated proctoring layer that institutions embed in their LMS (Moodle, Canvas, Open LMS, Blackboard) and the SUMADI Secure Browser desktop app that pairs with the SUMADI mobile app for synchronous monitoring. The AI stack focuses on facial recognition, typing-pattern biometrics, object detection in the webcam frame, and browser-activity tracking. Of those four signals, only browser-activity tracking even looks at the screen; the rest read webcam and keystroke timing. The browser-activity capture is implemented through standard Chrome / Chromium screen-capture APIs, which on Mac are wired to ScreenCaptureKit and which respect window-level exclusion.
Key points
- SUMADI was founded by Raul Rivera in 2019 and is now part of Open LMS, the LMS spun out from Laureate Education / Blackboard Open LMS.
- Strongest in LATAM and Spanish-speaking online education; processes millions of images per hour and supports seven languages per vendor docs.
- Two Mac surfaces: a browser-embedded proctoring layer and the SUMADI Secure Browser desktop client that pairs with the SUMADI mobile app.
- AI signals: face recognition, typing-pattern biometrics, object detection in the webcam, and browser tracking; only the last touches the screen.
- Webcam-side detection (looking-away, second face, phones in frame) is unaffected by the overlay because the webcam pipeline is separate.
How it works
┌── macOS host ─────────────────────────────────┐ │ ┌── Safari/Chrome ─────────────┐ │ │ │ ┌── proctoring extension ─┐ │ │ │ │ │ Browser screen-capture │ │ ← reads │ │ │ │ → exam recording │ │ visible │ │ │ └─────────────────────────┘ │ pixels │ │ └──────────────────────────────┘ │ │ │ │ ┌── LDBypass overlay ──────────┐ ← rendered │ │ │ ChatGPT / Claude / Ollama │ excluded │ │ └──────────────────────────────┘ from cap │ └───────────────────────────────────────────────┘
Compatibility on Mac
| macOS 14 Sonoma+ | Yes (required for window-privacy API) | ✓ |
| Apple Silicon (M1/M2/M3/M4) | Native arm64 build | ✓ |
| Intel Macs (2019+) | Universal binary supported | ✓ |
| SUMADI browser layer (Moodle/Canvas/Open LMS) | Chromium screen-capture honors privacy | ✓ |
| SUMADI Secure Browser (Mac) | User-space client, no kernel hook | ✓ |
| SUMADI mobile app (companion) | Phone camera is separate; close overlay during room scan | ~ |
Common questions
Does SUMADI detect overlay tools through its AI?
SUMADI AI looks at the recorded video and the keystroke pattern. An overlay that does not reach the recording cannot be analyzed by the AI.
Will the typing-pattern biometric flag me?
Typing biometrics watch your real keystrokes on the exam input. LDBypass does not synthesize input; you still type your own answers, so your typing pattern stays consistent.
Does SUMADI work with the mobile companion app for room scanning?
Yes. The mobile app uses your phone camera as a second view of the room. Close the overlay during the room scan.
Is the overlay invisible if a human reviewer escalates and watches my recording?
Yes. The reviewer watches the recording SUMADI captured. That recording was filtered through macOS compositor at capture time.
Which universities use SUMADI?
SUMADI is deployed by online and hybrid universities across LATAM and Spain, plus institutions inside the Open LMS / Laureate ecosystem.