Watson X-Ray Visualization

Brought in to give Watson a face. When IBM needed to explain their groundbreaking — and deeply complex — AI system to the world, they needed something that could make it tangible. I led the design and engineering of X-Ray, an interactive visualization framework that pulled back the curtain on how Watson actually thinks.

Built on a custom Three.js framework with D3-powered widgets, a scalable Java/CouchDB NoSQL backend, and wired directly to the Watson server room, X-Ray took any natural language question and dissected it in real time — breaking it down to its linguistic building blocks: nouns, verbs, geospatial markers, and every NLP signal Watson needed to truly understand intent. It then surfaced not just the answer, but the top 10 candidate answers, each with confidence scores and the evidence chain driving them.

The system was designed for scale across the IBM ecosystem — multiple users simultaneously interacting via voice, video, and multi-touch displays, with questions flowing in from IBM locations worldwide. Several iterations across Unity and Unreal led us to Three.js as the cutting-edge web-native solution that made that interconnected vision possible, trained against the full breadth of Wikipedia.

It was part demo, part tool, part proof — that AI could be understood, not just used.