Private temporary
cloud processing.
Statements use ZentraSnap’s own extraction pipeline, separate from Google Gemini.
- Private upload
- Extraction
- Cleanup
Private upload
Statements are uploaded to a private Cloudflare R2 bucket using short-lived authorised upload URLs.
Dedicated inference
Processing runs on Modal infrastructure using ZentraSnap’s document extraction model and OCR when required.
Short source-file retention
The source statement is normally deleted after the extraction job completes or fails. Temporary R2 objects also have a one-day lifecycle expiry rule.
No training without separate agreement
Production customer bank statements are not used to train ZentraSnap’s models unless the customer separately and expressly agrees.