Project Overview
Ava is a conversational AI healthcare assistant integrated into iMessage that helps University of Washington students navigate healthcare services by translating symptoms into structured clinical signals and routing them to the appropriate care destination.
Team Members: Alley Wu, Ani Ramadurai, Katharina Cheng, Mekias Kebede, Sofila Song
Process
Research: Interviewed UW students, validated that 94% use iMessage daily and moderate trust in AI healthcare. Created persona "Alex Chen" - international student unfamiliar with U.S. healthcare.
Design: Built symptom understanding system, care routing logic, and plain-language feedback. User testing revealed responses were too long - pivoted to add interactive elements (buttons/options) and requested features (cost estimation, in-network providers).
Development: Implemented pre-classification system, multi-language translation, and care routing with next steps.
Team Actions
Collectively chose iMessage platform based on usage data
Established ethical boundaries: "triage support" not "diagnosis"
Iteratively tested and refined based on user feedback
Developed privacy-first approach with minimal data collection
Key Takeaways
User needs evolve: Started with triage, pivoted to full care navigation
Ethics require proactive design: Clear language about AI limitations builds trust
Platform matters: Familiar interfaces (iMessage) reduce adoption friction
Iteration is essential: User testing drove major improvements in UX
Next Steps (Spring Quarter)
Improve onboarding process and interactive responses
Add insurance-based recommendations and cost estimates
Generate clinic visit scripts for users
Implement conversation encryption for privacy
Long-term: Expand to other universities, integrate with health center scheduling, track impact metrics.
Impact: Addresses healthcare equity for international students and low health-literacy users through accessible, multilingual AI assistance.
Links: Demo Video | Code Repo
Ava - Capstone Project
Team Token Tots | Sponsored by Photon | Winter-Spring 2026

