The business
International Biomedical is an Austin company that builds equipment for newborn intensive care. It is best known for transport incubators: the enclosed, heated, monitored units that carry premature and critically ill babies between hospitals by ambulance, helicopter and plane.
Medical devices are regulated, and every incubator has to be tested and certified before it leaves the building. That makes the test bench a bottleneck. The time a unit spends in testing limits how many can ship in a month.
The problem
The test machines write their results out as raw CSV files. Before an engineer could analyse a run, that output had to be reshaped into structured data the testing tools could read, and parts of the testing process were ripe for scripting.
On the software side, bugs were surfacing when branches were merged, and the product catalog the sales team works from could not be customized for every product.
What I built
- The parser. A data-parsing and formatting application in Python that reads machine-generated CSV files and writes structured JSON for device testing.
- The test framework. An automated testing framework that speeds up the analysis done during manufacturing.
- The integration. Python for the scripts, C++ where they had to plug into the existing systems, and Docker so development and test environments matched.
- The branching model. Restructured the team’s GitFlow charts on GitHub, with Jira for tracking work across teams.
- The catalog. Updated the product catalog interface the sales team uses so that every product could be customized in it.
From test bench to certified unit
- Test machine
- Raw CSV file
- Python parser
- Structured JSON
- Automated test framework
- Certified incubator
What changed
- 12%
less time spent testing, once the parser was in place.
- 10%
more incubators certified each month with the automated framework.
- 7%
fewer bugs found at code merge after the GitFlow rework.
