Coordinated AI triage of about 3,000 issues
I coordinated engineering, architecture, infrastructure and test experts on an AI defect-triage proof of concept that triaged about 3,000 issues on live Ultrasound data.
~3,000
issues triaged
- Without it
- Test failures would keep being triaged by hand, one record at a time.
- The call
- Prove triage on live data with people reviewing the AI's decisions, and settle architecture questions before building further.
- What I did
- Kept one requirements and reporting baseline for all the teams and co-wrote the architecture discovery questionnaire.
- What changed
- About 3,000 issues triaged, the first sprint demo delivered and a web app pulling live test data.