REDDIT · REDDIT// 2d agoINFRASTRUCTURE

Non-profit seeks free compute for 64M page OCR project

A non-profit organization is seeking free or subsidized compute resources to perform OCR on 64 million pages for a knowledge base project after exhausting credits on Vast.ai. The community suggested several avenues including major cloud provider grants (AWS, GCP, Azure), academic partnerships, and distributed computing frameworks like Petals or BOINC to handle the massive processing requirement.

// ANALYSIS

OCR at this scale is a massive infrastructure challenge, but the solutions are increasingly accessible via non-profit credits and optimized open-source stacks.

–AWS, Google, and Microsoft offer substantial grants for eligible non-profits, which is the most reliable path for large-scale compute needs.
–Distributed compute (BOINC, Petals) represents a viable community-driven alternative for non-sensitive data, leveraging volunteer hardware.
–Specialized "local" OCR stacks like Tesseract or PaddleOCR on cheap CPU clusters are far more cost-effective than LLM-based vision models at this volume.
–Academic partnerships can provide access to NSF-funded HPC clusters for high-impact non-profit work.

// TAGS

ocrnon-profitcloudopen-sourceinferencegpuself-hosted

DISCOVERED

2d ago

2026-04-10

PUBLISHED

2d ago

2026-04-10

RELEVANCE

7/ 10

AUTHOR

thereisnospooongeek