Builds · X-Ray Mobile Texas · 2026
Lab-order audit pipeline, every order read.
Every incoming lab order can now be audited automatically: 8,928 order emails read with zero extraction errors.
The problem
Lab orders arrived as faxed order forms in the company inbox. Reading them by hand was never going to happen, so nobody had turned order volume into what it meant for revenue.
What I built
A pipeline that pulls each order from Gmail, reads it with OCR, extracts the tests with an LLM, removes duplicates and reports.
- Parallel extraction made it eight times faster, and the output was verified identical to the serial run before it was trusted
- Raw order data deleted after each audit under the company's PHI rules, keeping only aggregates with no patient information
What it showed
The first audit showed that a small share of tests carried a large share of the revenue. That became a revenue model, built in about 30 minutes, for a decision nobody had been able to make before.
A later refresh showed faxed orders collapsing by about 95%. It wasn't a collapse: ordering had moved to the lab's portal. Combining both channels showed real volume had grown 5.6 times.
Results
Decisions about the lab partnership were made on what the orders actually showed, not on an assumption the data contradicted.
How I built it
With my agentic team, checking the business answer rather than just the code: when a number moved sharply, the pipeline was audited before anyone acted on it. HIPAA is part of the design: patient data minimized, access role-based at the API layer, and a person reviews before anything goes live.
One of six systems I built at X-Ray Mobile Texas, my family's mobile diagnostics company, during its turnaround. Shared with the owners' permission; patient, staff, client and vendor details are withheld. More operational case studies from my practice are at actional.us.