From scanned paper to structured grade
Not just OCR. Not just an LLM call. A full pipeline that reads the paper, understands the handwriting, applies your rubric, and returns auditable results with confidence scores.
Scanned PDF & Image Input
Accept multi-page PDFs, JPEGs, and PNGs from any scanner or mobile capture. Automatic deskew and quality checks.
Handwriting OCR
Extract handwritten text in English and regional languages. Handle cursive, print, and mixed scripts.
Math & Equations
Read handwritten mathematical expressions, solve against model answers, and award step-by-step partial credit.
Diagrams & Figures
Detect and interpret circuit diagrams, geometry, graphs, and labelled diagrams in handwritten responses.
Rubric-Based Scoring
Apply your marking scheme with criterion-level granularity. Support step marking and alternate valid approaches.
Confidence & Review
Every response gets a confidence score. Low-confidence results route to a human reviewer with evidence and reasoning.
How a handwritten grading job works
Send a scanned script. Receive question-level marks, evidence, and confidence. Async webhooks for batch processing.
Upload Scanned Scripts
Send scanned PDFs or images of answer booklets. We detect page count, orientation, and quality.
Question Segmentation
Detect and segment individual questions from scanned answer booklets. Map responses to the correct question numbers.
Handwriting Extraction
Extract handwritten text, equations, diagrams, and cross-outs. Reconstruct responses across multiple pages.
Rubric Evaluation
Apply your marking scheme. Allocate partial credit. Detect alternate valid approaches and award marks accordingly.
Confidence Scoring
Per-question confidence scores. Low-confidence results auto-routed to a human review queue with reasoning.
Structured Output
Question-level marks, evidence, feedback, and audit trail delivered via webhook or polling. Write back to your system.
Upload Scanned Scripts
Send scanned PDFs or images of answer booklets. We detect page count, orientation, and quality.
Question Segmentation
Detect and segment individual questions from scanned answer booklets. Map responses to the correct question numbers.
Handwriting Extraction
Extract handwritten text, equations, diagrams, and cross-outs. Reconstruct responses across multiple pages.
Rubric Evaluation
Apply your marking scheme. Allocate partial credit. Detect alternate valid approaches and award marks accordingly.
Confidence Scoring
Per-question confidence scores. Low-confidence results auto-routed to a human review queue with reasoning.
Structured Output
Question-level marks, evidence, feedback, and audit trail delivered via webhook or polling. Write back to your system.
Google Classroom
Canvas
Blackboard
MoodleHandwritten Grading API FAQ
What input formats does the handwritten grading API accept?
The API accepts multi-page PDFs, JPEGs, and PNGs from any scanner or mobile capture. It automatically detects page count, orientation, and quality, and performs deskew and quality checks.
How does the API handle messy or unclear handwriting?
GradeLab uses multiple vision-language models to read handwriting, not a single OCR engine. Cross-outs, smudges, and overlapping text are detected and handled. Every response gets a confidence score, and low-confidence results are automatically routed to a human reviewer with evidence and reasoning.
Can the API grade mathematical equations and diagrams?
Yes. The API reads handwritten mathematical expressions, solves them against model answers, and awards step-by-step partial credit. It also detects and interprets circuit diagrams, geometry, graphs, and labelled figures.
What is the batch size limit per API call?
You can send up to 500 submissions per API call. Results are delivered asynchronously via signed webhooks or you can poll the status endpoint. Idempotency keys ensure safe retries and automatic deduplication.
How does human review work with the API?
Low-confidence results are automatically routed to a review queue. Reviewers see the AI marks, evidence, and reasoning before deciding to approve, override, or adjust. Every override is timestamped and retained in the audit trail.
Does the API work with regional languages?
Yes. The handwriting OCR supports English and regional languages including Hindi, Tamil, Telugu, Bengali, and more. It handles cursive, print, and mixed scripts.
All GradeLab API products
AI Grading API
The broad API hub. Add AI grading to any education platform.
Assessment & Scoring API
Score essays, short answers, STEM, and mixed-format assessments.
Embedded AI Grading
White-label AI grading UI inside your product. Your brand.
ERP AI Grading
Add AI grading to your education ERP examination module.
AI for On-Screen Marking
Turn your OSM into AI-assisted evaluation. Three modes.
Subjective Answer Grading API
Step marking, partial credit, and alternate valid approaches.
Exam Grading API
Grade paper and digital exams through one API.
LMS AI Grading
Add AI grading to your LMS via LTI, AGS, and OneRoster.
Send a scanned script. Receive a structured grade.
Book a demo and see the API grade a handwritten exam in real time. We'll walk you through integration architecture, deployment options, and pricing for your volume.
Book a Demo