RocketGrades
Grading is the work teachers like least and do most. RocketGrades automates the first pass: reading submissions in any format and returning structured, rubric-aligned feedback.
Teachers lose hours to repetitive grading and feedback, and students wait days for it. The opportunity was to compress that loop without replacing the teacher's judgment.
RocketGrades gives teachers and students separate role-based portals, accepts submissions in PDF, Word, or text, and runs an AI first pass that produces structured feedback the teacher can review and adjust.
Auth & roles
Auth0 handles authentication; JWT-based role-based access control routes teachers and students to their own dashboards with the right permissions.
Angular portals
An Angular 19 + TypeScript front end (Tailwind/DaisyUI) drives both portals: class and assignment management for teachers, submission and feedback views for students.
Submission storage
Uploaded files land in AWS S3; metadata and relationships (classes, assignments, submissions) persist in MongoDB through Mongoose schemas.
AI grading
A Node/Express service uses LangChain + Mistral AI to parse PDF/Word/text submissions and produce rubric-aligned, structured feedback rather than just a number.
LMS-aware chat
A chat assistant carries LMS context (the class, the assignment) so its answers are specific to the course, not generic Q&A.
Lean on a hosted model, not a custom one
Integrating Mistral AI through LangChain for document analysis meant no training pipeline to build or maintain. The right call for a hackathon, and it kept the focus on the grading workflow and UX.
Role-based access from day one
Teachers and students see fundamentally different apps. Building RBAC on Auth0 + JWT up front meant the two portals could diverge safely instead of bolting permissions on later.
Structured feedback over raw scores
The AI returns rubric-aligned feedback the teacher can edit, keeping a human in the loop. The model accelerates the first pass; it doesn't own the grade.
Built with