Documentation
RavenGrader reads handwritten STEM work and proposes grades against a rubric built from your answer key. You review and approve every score before students see it.
What it does
The unit of work is an assignment (an exam, quiz, or problem set) inside a course. For each assignment you provide an answer key; RavenGrader turns it into an editable rubric, then grades submitted work against that rubric question by question.
- Handwriting. Scanned or photographed pages, PDFs, and typed text. Diagrams and circuits are read rather than skipped.
- Per-question scoring with written feedback, not a single overall mark.
- Two grading tracks: AI grading with review, or manual grading by question. See Grading.
- Class analytics: question averages, score distributions, and common mistakes.
The shape of a term
- Create a course, then add an assignment to it.
- Upload the answer key. RavenGrader generates a rubric you can edit.
- Collect student work: upload it yourself, share a join code, or receive it through Canvas.
- Grade, then review and approve.
- Return per-question feedback reports, or export grades to your LMS.
Quickstart walks through one assignment end to end.
You approve every grade
AI output is a draft. Nothing reaches a student until an instructor approves it, and you can change any score or comment before or after approval. Grades on a student’s report page render live, so a later point edit shows up without regenerating anything.
Where to go next
- Quickstart: your first graded assignment
- Grading: how AI grading works, the manual track, regrading
- Collecting work: uploads, join codes, student reports
- Canvas (LTI): installing and launching from Canvas
- FAQ: credits, accounts, roles, data, accuracy
For privacy and student-data questions, see the Privacy Policy (Section 7 covers student data and FERPA). Security researchers should read our vulnerability disclosure policy.