Problem
Clinics and care teams receive care-event records from multiple intake channels. Before any downstream automation or reporting, someone needs to confirm required fields are present, flag missing risk data, and route ambiguous records for human review instead of silently dropping or misclassifying them.
Data boundary
Synthetic-only sample records, generated for this lab. No real patient data, no PHI, no connection to any clinical system.
Pipeline
Ingest synthetic care-event record → validate required fields → flag missing risk data → route to ACCEPT, QUARANTINE, or review → produce a review packet.
Validation rules
| Field | Rule | Sample value | Outcome |
|---|---|---|---|
| patient_ref_id | Required, synthetic identifier only | SYN-00214 | ACCEPT |
| risk_flag | Required for triage routing | (missing) | QUARANTINE — missing risk field |
| event_timestamp | Required, ISO-8601 | 2026-06-30T14:02:00Z | ACCEPT |
| source_channel | Must match an approved intake channel | unknown-fax-gateway | QUARANTINE — unapproved source |
Output
A review packet per record: validation outcome, flagged fields, and routing decision, ready for care-team review before any further automation.
Failure modes
- Missing risk field → quarantined, never silently accepted.
- Unrecognized source channel → quarantined for owner review.
- Malformed timestamp → rejected with a structured reason, not a raw parser error.
What this proves
Schema validation, quarantine routing, and audit-ready review-packet design applied to a healthcare-shaped intake problem.
What it does not claim
This lab uses a HIPAA-aware design pattern (schema validation, quarantine, audit trail) demonstrated on synthetic data. It is not a HIPAA-compliant system, has not undergone legal or compliance review, and has never processed real patient data.
Suggested questions
- Hiring manager: "Show me the health data intake pattern."
- Buyer: "How would you validate messy intake data before automation?"