Work

Built: Stateful interruption-and-resume task model, new IA for context-preserving workflows, A/B testing platform (Claude Code), design system components across lab ops products.

Scientists were pulled mid-workflow for priority samples. The system had no concept of “resume.” Invisible in logs, obvious when you stand next to someone working.

Read full case study →

Built: Patient-controlled result delivery (not push — patients choose when to see genetic test results), educational context layers, complete interaction model and emotional pacing.

Patients didn’t want easier access. They wanted control over timing. Genetic testing anxiety means people need to choose when to receive life-changing info.

Built: Design principles for health tools that respect embodied experience. Concrete patterns for accommodating ambiguity. Systematic review + MARS evaluation + 40-person diary studies → framework that changes how I build everything.

The most clinically top-rated menopause apps were the least used. Women abandoned tools that optimized menopause. The gap between “evidence-based” and “evidence-felt” was the core finding.

Built: Dual-entry IA (patient-centric for MDs, dataset-centric for bioinformaticians), pipeline selection interface, full intake flow, personas, style guide, UX packet.

MDs thought in patients. Bioinformaticians thought in datasets. The upload flow assumed one model. I built two front doors into the same data.

See live product ↗

Built: Complete IA restructure — from agency org charts to child-needs navigation (health, legal, placement, education). New interface, 4 personas, card sort-validated structure.

The portal reflected bureaucracy, not user tasks. Foster children’s health data was split across organizational boundaries that meant nothing to social workers.