Led development of AI-powered symptom checker that optimized urgent healthcare services and reduced unnecessary ER admissions.
Emergency rooms were overwhelmed with non-urgent cases, while patients struggled to determine appropriate care levels. This created poor experiences and inflated healthcare costs.
42% of ER visits were classified as non-urgent
Average wait time: 3.5 hours for urgent care
User research showed 78% uncertainty about when to seek emergency care
Conducted interviews with 30 patients and 12 healthcare providers
Found that users wanted guidance, not diagnosis
Identified trust as the critical factor for adoption
Legal and accuracy concerns, plus users didn't trust black-box AI for health
Balanced user trust with intelligent routing – transparent logic with AI optimization
Added cost and complexity; users wanted quick guidance, not always a consultation
Focused on common symptoms (80% of cases) vs comprehensive coverage
Built native integration vs iframe – took longer but better UX
Invested in content accuracy over fancy UI animations
Partnered with clinical team to validate decision trees
Ran A/B tests on question phrasing for clarity
Coordinated launch with customer support for potential edge cases
Built comprehensive analytics to track patient outcomes
Would have soft-launched to a smaller user segment first
Should have set up better feedback loops with clinical staff
Would advocate for more time on edge case handling