Inspiration

Manual grading of academic assessments is a time-intensive process that can take educators 15-20 minutes per paper, while also introducing inconsistencies in evaluation. Assessly leverages advanced AI models to automate this workflow—reducing grading time by 85% while ensuring standardized, fair, and transparent feedback that benefits both educators and students

What we built

Assessly is an AI-powered grading platform that automates academic assessment evaluation. Built with a FastAPI backend and Next.js frontend, it features:

Multi-modal OCR: GPT-4 Vision for handwritten math, Google Cloud Vision for text Intelligent Grading Engine: Subject-specific AI models (GPT-4o for STEM, Gemini 2.0 Flash for conceptual subjects

Real-time Progress Tracking: WebSocket + polling for live grading status updates

Visual Annotation Canvas: Interactive PDF viewer with AI-generated feedback overlays

Smart Data Extraction: Automatically parses student details, class sections, and subjects from worksheets

Dynamic Score Management: Live updates across UI components when scores are adjusted

Organized Dashboard: Grades automatically categorized by class section and subject"

Problem it solves

Assessly eliminates the time-consuming, inconsistent manual grading process by using AI to evaluate academic assessments in minutes with standardized, transparent feedback.

How AI is used

Assessly uses AI across three critical stages:

OCR & Text Extraction: GPT-4 Vision reads handwritten math (equations, diagrams), while Google Cloud Vision processes typed text from student worksheets

Rubric Parsing: GPT-4o converts uploaded rubric PDFs into structured JSON, extracting point allocations, grading criteria, and partial credit rules

Intelligent Grading: Subject-specific models evaluate student work—GPT-4o for mathematical reasoning and STEM subjects, Gemini 2.0 Flash for conceptual subjects (English, History, Science)—comparing answers against rubric criteria and generating detailed, question-by-question feedback with scores and explanations."

Target users

Primary Users: K-12 Teachers & University Professors grading 30-100+ assessments weekly across subjects like mathematics, computer science, physics, and conceptual courses

Teaching Assistants (TAs) handling bulk grading for large lecture courses

Online Education Platforms needing scalable, automated grading infrastructure

Secondary Users: Students receiving faster, more detailed feedback on their work Department Heads tracking performance analytics across sections and subjects

what makes Assessly unique.

1. Multi-Modal Intelligence Unlike single-model solutions, we use subject-specific AI: GPT-4o for precise mathematical reasoning and Gemini 2.0 Flash for conceptual evaluation—optimizing both accuracy and cost.

2.Visual Annotation Canvas We don't just return scores. Our interactive PDF editor displays color-coded annotations, feedback boxes, checkmarks, and error markers directly on student work—mimicking how teachers naturally grade but with AI speed.

3. Smart Extraction + Auto-Organization Assessly automatically reads student details, class sections, and subjects from worksheets, then organizes grades into a searchable dashboard by class/subject—no manual data entry required.

4. Full Transparency + Control Educators can edit any AI-generated score or feedback with instant recalculation across the entire UI. It's AI-assisted, not AI-dictate

What's next for Assesly

LMS Integration Direct plugins for Canvas, Blackboard, and Google Classroom—grade within existing workflows without file downloads.

Adaptive Learning Feedback AI that suggests personalized practice problems based on grading patterns and common student errors.

Built With

Share this project:

Updates