Nursa: AI TB Verification Model

Problem / Challenge

Nursa's credentialing process relied on a costly third-party vendor for TB test verification. Manual review was expensive, slow, and introduced data quality issues that blocked clinician onboarding.

Solution / Approach

Built an AI-powered TB test verification model trained on real user data. Collaborated with developers to extract training samples, run model training, and deploy to production. Also audited Segment event tracking to close critical analytics data gaps.

Results / Outcomes

$30K in annual cost savings; improved data collection quality; in-house model replaced vendor entirely; product analytics materially strengthened.

Tech Stack

Lessons Learned

“Replacing a vendor with an in-house model requires more than technical accuracy — you need to match the vendor's reliability guarantees and build internal confidence before you can cut the contract. Tracking edge cases early is critical.”