Clinical document watcher

The problem

Provider endpoints across many sites generated signed clinical documents that needed to land in a central store with chain-of-custody logging. The legacy approach scheduled FTP polls and manual uploads; documents were lost, duplicated, or arrived without provenance. There was no self-healing when an endpoint went offline. Compliance review surfaced the gap, and the team needed a fix that did not require staff at every endpoint.

The approach

We built a fleet of .NET file-system watchers running unattended at every endpoint. Each watcher detects new signed documents, classifies them, OCRs where needed, and pushes to a central store with cryptographic chain-of-custody logging. When the network or central store is unavailable, the agent buffers locally with checkpoint markers; when it recovers, it resumes without duplicates. Diagnostics ship from each agent to a central monitor so ops sees the fleet state without polling each site.

Stack and engineering choices

• .NET FileSystemWatcher agents
• OCR fallback for scanned docs
• S3-style central object store
• Chain-of-custody hashing
• Checkpoint-based resume
• Self-healing on outage
• Per-agent diagnostics

Outcome

Documents arrive at the central store in near-real-time with provable provenance. Endpoint outages no longer cause data loss; recovery is automatic when the network returns. The ops team monitors the agent fleet from one dashboard instead of polling each site, and chain-of-custody questions resolve from the audit log without forensics.