MOM's Bridge 소개

Learning Journey

Final Submission Package
What We Learned (100 words)

We learned how to integrate Google Gemini AI's multimodal capabilities into a real-time medical translation system. Our biggest technical challenge was getting structured JSON output consistently -- we iterated through multiple prompt versions before settling on explicit JSON-only instructions with word limits. We learned about data privacy requirements and implemented account deletion with cascade deletes, a translated privacy policy, and data export. Building the i18n system for 11 languages taught us about fallback mechanisms and Accept-Language header parsing. Implementing the daily tracking feature taught us about ACOG clinical guidelines and the importance of deterministic safety logic over AI for health-critical decisions.

Word count: ~99
How We Validated Information

We validated pregnancy health data against CDC PRAMS (Pregnancy Risk Assessment Monitoring System), which is publicly available and federally maintained. Pregnancy week milestones were cross-referenced with established medical sources. Daily tracking risk thresholds follow published ACOG guidelines. For translation quality, we built an English fallback mechanism -- missing translations gracefully fall back rather than showing broken content. We used Gemini's structured JSON output mode to ensure AI responses were consistently parseable, with explicit length constraints.

Bibliography