VeriQAi publishes free, open-source, privacy-first AI tools for educators. Every tool runs entirely in your browser: no accounts, no subscriptions, no student data sent to our servers. Two tools are available today:

GradeBridge is a system for creating structured engineering course assignments with AI autograding via Gradescope. It replaces ad-hoc PDF submissions with a structured workflow: students submit an encrypted ZIP archive, and a Docker-based autograder (running in your own Gradescope account, calling the Anthropic Claude API) scores answers automatically against your rubric.

GradeBridge does not use Gradescope's built-in AI features. It does not use Gradescope PDF templates or template regions. No student data passes through any VeriQAi server at any point.

flowchart TD
    A["📄 Instructor's existing materials\nWord · PDF · Markdown"]
    A -->|"Claude Code reads source\ntwo-phase: confirm objectives → generate rubrics"| B["📝 Structured assignment .md file"]
    B -->|"Instructor imports & reviews"| C["🖥️ Assignment Maker\nbrowser app, no install, no account"]
    C --> D["📬 Student assignment file\ndistributed via Canvas"]
    C --> E["🔒 Confidential grader document\nHTML, for TAs"]
    D -->|"Student loads in browser"| F["🖥️ Student Submission App\nno install · no account"]
    F -->|"Student fills answers, clicks Download"| G["📦 Encrypted ZIP archive\nAES-256-GCM"]
    G -->|"Student submits to Gradescope"| H["⚙️ GradeBridge Docker autograder\nPython · calls Anthropic Claude API"]
    H --> I["✅ AI-graded questions scored automatically\n🔍 TA-review questions flagged for human review"]

1. Instructor: Run Claude Code against your existing lab manual or project description (Word, PDF, or Markdown; no reformatting needed). Claude Code generates a structured assignment .md file in two phases: first you confirm the learning objectives, then rubrics are written.
2. Instructor: Import the .md into the Assignment Maker browser app. Review and adjust questions, point values, and rubrics. Export. This produces the student assignment file (distribute via Canvas), a PDF, and a confidential grader document (HTML, for TAs).
3. Student: Open the Student Submission App in a browser (no account, no install). Load the assignment file, fill in answers, click Download for Gradescope. Submit the single ZIP archive to Gradescope.
4. Gradescope: GradeBridge's Docker autograder runs. It calls the Anthropic Claude API directly, scores AI-graded questions automatically, and flags TA-review questions for human review.

• A Gradescope account with a course
• An Anthropic API key (used by the Docker autograder on Gradescope)
• Claude Code (for assignment generation from your existing materials)

• VeriQAi/VeriQAI-GradeBridge: system docs and CCAssignmentMaker tools
• VeriQAi/GradeBridge-Assignment-Maker: Assignment Maker app
• VeriQAi/GradeBridge-Student-Submission: Student Submission app

Creating a single MATLAB Grader problem from scratch means writing four separate artifacts: problem description, reference solution, learner template, and test cases (each with MATLAB Grader's exacting syntax requirements). The test cases alone (guard conditions, diagnostics, R2025b API) can take an experienced instructor an hour or more per problem.

Enter a learning objective. The app proposes calibrated problems across Easy / Medium / Hard difficulty for Script or Function problem types. Select the ones you want: all four artifacts are generated in sequence with live review at each step. Download as a named ZIP ready to paste directly into MATLAB Grader.

Key facts:

• Requires your own Anthropic API key (get one at console.anthropic.com)
• Generating and fully developing 4 problems costs approximately $0.05-$0.10 with Claude Sonnet
• R2025b-compliant MATLAB Grader syntax out of the box
• No signup · MIT License · Runs entirely in your browser

Repository: VeriQAi/MatlabGraderProblemGenerator

We are educators and engineers who build tools to remove friction from AI-assisted teaching and assessment. Our tools work alongside the platforms you already use (Gradescope, MATLAB Grader, Canvas) rather than replacing them.

All tools are MIT-licensed, free forever, and designed so that your data and your students' data never leave your machine.

Report an issue or request a feature