bilal@dev:~/projects/gradeloop$_
GradeLoop
AI-driven coding LMS with intelligent grading & feedback
Research project: an AI-integrated coding LMS that combines autograding, Socratic tutoring, code-clone detection, keystroke analytics, and AI viva assessment across an event-driven microservice platform.
Papers analyzed
50+
Team
4
Milestones
9
Publications
2
> Problem
Programming courses still grade late and shallowly — students get binary pass/fail or full solutions, while instructors lack tools for plagiarism signals, oral assessment at scale, and formative feedback that teaches instead of spoiling answers.
> Approach
Designed a polyglot microservice stack: Next.js for the LMS UI, Go/Fiber services for IAM, academic, and assessment orchestration, Python/FastAPI for CIPAS similarity + AI pipelines and IVAS voice vivas, with RabbitMQ events, PostgreSQL per service, and MinIO for submissions. Feedback is Socratic — hints and rubric commentary guide students without dumping the solution.
== [ GALLERY ] ==
> Challenges
Coordinating at-least-once grading workers under bursty submission loads, sandboxing untrusted student code, keeping AI hints pedagogically safe (no answer leaks), and stitching realtime IVAS voice sessions with speaker verification into the same LMS flow.
> Learnings
Research constraints become product constraints: clone detection, typing signals, and viva integrity only matter if the UX stays fast enough for a live classroom. Event-driven boundaries let AI pipelines evolve without blocking the interactive path.