A client-management platform plus a claims-collection automation for a US medical-billing company serving Hospice and Home Health providers — it gathers the claims itself, flags the broken ones on arrival, and keeps every byte on HIPAA-compliant cloud.
A US company does medical billing for Hospice and Home Health providers. It needed one internal platform to run the whole operation — a single place for its client organizations, their patients and physicians, the claims and the errors, all of it validated and re-checked, on reliable HIPAA-compliant infrastructure.
One of the heaviest daily jobs was getting claim statuses. Operators worked by hand through a terminal session: for each client they connected, entered a login and password, switched between screens and pulled the data out manually. The terminal is slow — extracting data for a single client took several minutes, and a full pass across all clients stretched to tens of minutes (around 35 and more). Every single time.
BillingHub — one internal platform, built from scratch, end to end. A single system that holds the client organizations (Hospice / Home Health), their patients and physicians, the claims and the errors — and validates everything several times over. The claims-collection automation is one standout part of a much larger whole.
A single registry of client organizations (Hospice / Home Health) with business-specific custom fields, service statuses (active / suspended / terminated) and a full change history per client — who changed what and when, down to the individual field. Bulk import of clients is supported.
A flexible role model: for each role you configure exactly which fields an employee can view and which they can edit, and which slice of clients they reach. Access to sections — the claims catalog, client management, record deletion — is just as granular, so each person gets exactly what their work needs and nothing more.
The system checks entered data automatically. ICD-10 codes are verified against icd10data — a code must match the diagnosis and the patient’s gender and must not be on an internal blocklist. Treating physicians are checked against CMS (data.cms.gov) for the required authorizations, catching errors in codes and patient data early.
Deployed on HIPAA-compliant Google Cloud (covered by a BAA) with managed Cloud SQL, Secret Manager and encryption. Access runs on signed, short-lived tokens with one active session per user; sign-in adds TOTP two-factor. Every mutating action is recorded in an audit log, passwords are stored only as hashes, and the automation talks to the platform over a protected channel.
Pulling claim statuses was the worst of the daily grind — so it became the platform’s most powerful part. The old way: an operator opened a legacy terminal, signed in client by client, paged through screens and copied records out by hand. Minutes per client, around 35 and more for a full pass — every single time.
Now a scheduled automation does it end to end: it connects to the external claim-status system, walks every active client and loads thousands of claims in a single pass — managing which clients to collect, watching credential validity and logging every session. Classification runs alongside, so problem claims are flagged on arrival and operators open a ready-made list instead of re-reading everything.
See it below. By default a cursor works the terminal the slow, manual way — one patient at a time. Hit “Run automation” and the same work runs hands-off: pages flip and records stream into the store.
Reconstruction for this case study — all data is fictional; the real system is under NDA.
This project was delivered under an NDA, so what you see here is intentionally incomplete — the interface is reworked for this write-up (not a real screenshot), some functionality is omitted, and certain automation details are withheld at the client’s request. None of it bends the rules.
The work was done under a signed NDA and an established data-handling protocol. Internal details, client names and anything that could identify a patient are deliberately left out of this case.
No real sensitive or personal medical data was ever disclosed or used outside the protected environment — everything is stored and transmitted encrypted. All data shown in the demo is entirely fictional.
The claims-collection automation runs strictly within the standard software and interfaces the authorized agency itself provides — no hacking, security bypasses, reverse engineering or unauthorized access, and nothing that could disrupt or block government systems.