For partners
Verified, anonymized pages of real student work with ground-truth transcriptions, licensed with explicit consent and formatted for ML pipelines.
"subject": "Linear Algebra",
"image": "page_0481.png",
"transcription": "det(A - λI) = 0 ...",
"answer": "λ = 1, λ = 2",
"verified": true,
"consent": true,
"identity": null
Every record pairs a high-resolution scan of real handwritten work with a verified ground-truth transcription, tagged for filtering.
Records cover dense math notation, chemical structures, diagrams, margin notes, and crossed-out work: the cases where production OCR actually breaks.
Students opt in to share handwritten practice pages through the Spaide app, and get paid for every accepted submission.
Each page is checked for quality and transcription accuracy. Personal data is stripped before any record enters the catalog.
Accepted pages are organized into structured datasets with rich metadata, ready for licensing and export.
Download datasets in JSONL, Parquet, or HuggingFace format and drop them straight into training or evaluation pipelines.
Provenance you can put in front of legal. Every record carries explicit consent, and identity never enters the pipeline.
Students are paid contributors, not scraped sources. Licensing terms, consent records, and anonymization steps are documented for every dataset we ship.