Mask-Vault
Privacy-first document system using local OCR, NLP, and Document AI for purpose-aware redaction and secure sharing.
Inspect the architecture signal.
Privacy boundary is explicit: the supplied project requires local, open-weight processing with no external APIs.
Document
Document enters the privacy-critical local processing boundary.
Problem
Use OCR, NLP, and Document AI for purpose-aware PII redaction and secure document sharing.
Constraints
No external APIs for the privacy-critical AI pipeline.
All components run locally using open-weight models.
Privacy-by-design is a system requirement rather than a presentation claim.
Decisions / trade-offs
Keep the core document-processing pipeline local instead of sending sensitive document contents to external AI services.
Ground the design in research across PII redaction, differential privacy, federated learning, and document layout understanding.
What broke / changed
Project is currently in progress.
Verified project facts
Research references include Flower/FedNLP and the LayoutLM series.