OmniGrades

My logo

Inspiration In 4th grade, my teacher shared a story that stayed with me: because of relentless, tedious grading, she had to bring stacks of assignments to a wedding just to keep up. Seeing educators lose their personal time to manual grading motivated me to find a solution. As a student, I also saw the flip side—delayed feedback hurts learning. I wanted to build a tool that gives teachers their time back while giving students immediate, actionable insights.

What It Does OmniGrades is an AI-powered grading and classroom management platform designed to automate evaluation, streamline assignment creation, and enhance student outcomes:

Multimodal Auto-Grading: Analyzes student submissions—including handwritten tests and rubrics—using Gemini to deliver fast, objective scores.

Instant Actionable Feedback: Provides personalized coaching and explanations to help students understand where they can improve.

Role-Based Workflows: Tailored dashboards and permissions for Admins, Teachers, and Students to manage classrooms, rosters, and assignments in one cohesive ecosystem.

How We Built It AI Core: Integrated Google Gemini to handle complex rubric comprehension, OCR test evaluation, and conversational student coaching.

Development Workflow: Leveraged Google Antigravity to orchestrate AI-assisted development, streamline architecture planning, and interface with Google Cloud infrastructure.

Full-Stack Architecture: Built a modern frontend with React and styled components for an intuitive user experience, backed by scalable backend services and Cloudflare/Cloud infrastructure.

Challenges We Ran Into Our initial prototype used vanilla HTML/CSS and CommonJS, which quickly became unmaintainable as feature complexity grew. Refactoring the entire codebase into a modern React SPA with ES modules required rewriting core state management, handling complex asynchronous API pipelines, and resolving regression bugs under a tight summer development schedule. This migration dramatically improved frontend performance, component reusability, and overall UI responsiveness.

Accomplishments We're Proud Of Successfully deploying the core multimodal auto-grader with accurate rubric alignment.

Implementing a complete three-tier role-based access control system (Admin, Teacher, Student).

Mastering cloud deployment workflows, environment configuration, and API orchestration within Google Cloud and Antigravity.

What We Learned Iterative Prompt Engineering: Modularizing tasks—feeding focused, single-objective context to Gemini rather than monolithic requests—yields far more robust code generation and debugging results.

Production Architecture: The importance of starting with scalable modular architectures (TypeScript/React/ESM) early in the development lifecycle.

What's Next for OmniGrades LMS Integrations: Develop direct API integrations and LTI compliance for Google Classroom and Canvas, allowing seamless one-click grading within existing school workflows.

Comprehensive AI LMS: Expand OmniGrades beyond grading into a full-scale, AI-driven Learning Management System featuring automated quiz generation, predictive student performance analytics, and interactive multimodal study assistants.