Central CRM
Hospices, patients, doctors and orders in one place. Flexible intake (files, tables, free text) — the system works out which field is which. Every operator works only here; no more “read it here, paste it there.”
← All cases Core Validation Legacy Reporting
Automation
A Medicare-billing platform that runs itself
A multi-service platform that takes the entire Medicare hospice billing routine off people’s hands — it validates, assembles, submits and reconciles the data itself.
The problem
Medicare hospice billing is one of the gnarliest corners of US healthcare: tangled rules, non-standard forms, and government systems you reach over protocols from the 90s. Many hospices can’t handle it themselves and hand it to specialist firms.
Inside the client, dozens of people did the same thing every day by hand — log into government systems, gather data, copy fields between emails, fill long forms. One wrong field means a rejected claim, a fine, an investigation. In Medicare, the cost of a mistake is real money.
It started small — “build us an ICD-10 code database.” Once I understood the business, I proposed closing the routine step by step across the whole chain. It has been growing ever since.
What I built
A platform of four interconnected services around one central system — each updatable on its own, none able to break its neighbours.
Multi-layer checks
Before anything ships, the system checks itself from several angles: does the diagnosis code fit this patient, are the codes mutually compatible, and — via an external government API — was this doctor even allowed to order this care.
BlueZone terminal automation
One government system speaks a 90s terminal protocol. I built an abstraction over BlueZone that types data, watches the cursor and the screen, saves results and grabs confirmations — with a re-check at every step.
Filing & report parsing
Submits the billing files, collects the returned claim IDs, pulls the key numbers out of the response and builds an email summary for the client. Plus a separate collector that hands accountants a ready report instead of a manual website crawl.
What made it hard
- Cost of error = real money. Every service re-checks that it did exactly what it intended — not just that it did something.
- Legacy protocols. BlueZone isn’t an HTTP call — it’s terminal emulation with cursor positions and screen states. I built an abstraction so the automation ignores the terminal’s quirks.
- Real medical rules. Code compatibility, doctor permissions, order types — not if/else, but formalized domain knowledge gathered with the client.
- Sensitive data throughout — data isolation, careful logging, access roles, separate environments.
The result
- Grew from a one-off ICD-10 task into a full platform covering the whole billing chain.
- Staff moved off mechanical work onto harder problems — the system does the busywork.
- The client keeps reallocating the freed-up capacity into growth — the best proof it works.
- Still live and expanding — a long-term, high-trust contract.
What I took awayAll cases →Think like a partner, not a contractor — the client arrived with one small task, but behind it sat a huge layer of pain he hadn’t even articulated. And validation as an architectural principle from day one, not “let’s check it later”: in high-risk domains, multi-layer error protection belongs in the skeleton.