How It Works

One assignment. One seamless loop.

From linking an assignment in your LMS to reviewing grades, diffs and session replays: here's the journey every submission takes through GradeLoop.

01
PHASE 01

Assignment created & linked

The lecturer builds a coding assignment in GradeLoop with rubric, starter code and constraints, then links it to their existing LMS through LTI 1.3. No new portals for anyone.

System Walkthroughs:
LTI 1.3Rubric Setup
02
PHASE 02

Student launches from the LMS

One click in the LMS opens GradeLoop’s sandboxed browser IDE, with nothing to install. Biometric authentication starts a fully tracked coding session before the first keystroke.

System Walkthroughs:
Cloud IDEBiometric Auth
03
PHASE 03

Build with Socratic support

When a student gets stuck, the Socratic tutor responds with guiding questions instead of answers. Every struggle signal, including failed runs, long stalls and hint requests, is logged.

System Walkthroughs:
Socratic TutorStruggle Tracking
04
PHASE 04

Competency viva

On submission, an oral assessment is generated from the student’s own code and their responses are evaluated automatically, proving they understand what they wrote.

System Walkthroughs:
Automated VivaAdaptive Questions
05
PHASE 05

Automated grading & feedback

The Grading Engine reasons over logic, correctness and style to award partial credit like an expert, returning automated feedback and a grade while the Integrity Engine clusters every submission by Type 1–4 similarity.

Automated FeedbackType 1–4 Clusters
06
PHASE 06

Instructor review & oversight

Review and override any grade, compare similarity clusters side-by-side, replay full coding sessions and see exactly where each student struggled, all from one dashboard.

System Walkthroughs:
Grade OverrideSide-by-Side DiffSession Playback
Trust & Safety

AI you can actually trust with assessments.

Academic integrity isn't a feature flag; it's the foundation the whole loop runs on.

Isolated execution

Every submission runs in its own containerized sandbox. No shared state, no escape routes, no side effects.

Biometric liveness

Face verification and liveness checks tie every session to a real, present student, not a shared login.

Full provenance trail

From first keystroke to final submit, every edit is logged. Grades come with evidence, not assumptions.

See GradeLoop grade live.

Watch a real submission travel the full AI pipeline, from raw code to rubric score and Socratic feedback, in under two seconds.