Rae is an AI product builder and interaction designer who turns ambiguous product questions into working, testable interfaces. She combines 8+ years of enterprise fintech consulting and systems engineering with an MDes in Interaction Design from California College of the Arts.
Based in San Francisco · MDes, CCA (expected August 2026) · F-1, STEM OPT eligible
A spatial conversation browser that evolved from a desktop sphere into a VR reading wall. Research and four exploratory walkthroughs led to an orient → focus → read → return interaction model, inspectable placement, seated reading, and explicit comfort questions.
Signals: interaction design, design research, spatial UI, WebXR, Three.js, Claude API.
An iOS personal-tracking prototype shaped by literature review, paper system sketches, and information-architecture redesign. The case study explains six decisions around uncertainty, daily use, personal context, constrained AI, privacy, and inspectable reports.
Signals: health UX, SwiftUI, HealthKit, on-device AI, evidence boundaries.
A voice-first language practice product where coaching had to be useful without becoming accent policing. The prototype uses one actionable upgrade per turn, mode-specific prompt rules, and a clear separation between model output and learner control.
Signals: conversational UX, AI product engineering, prompt systems, audio interaction.
Enterprise consulting and product work for a large Korean cooperative bank: prioritizing AI use cases, authoring a strategy report, and building cloud-hosted pilots with on-premise transition architecture.
Signals: product strategy, RAG, enterprise delivery, financial systems.
A multi-agent video intelligence pipeline that turns unstructured footage into searchable temporal findings and cited explanations.
Signals: agent orchestration, video analysis, retrieval, technical prototyping.
Rae writes a product question and evidence boundary before opening an AI tool, prototypes to expose assumptions, and audits the result for unsupported claims. Her portfolio documents not only shipped interfaces but also rejected alternatives, trade-offs, and what still needs testing.