Designed the public playground for IBM's Granite 4.0 models: the primary activation surface for the launch.
2.6× daily prompts · 3× unique users vs Granite 3.3
Not just pretty things: I care that it reads at a glance, works without friction, and solves the problem it was built for. End-to-end: research, interaction design, production code. Building with AI, I ensure consistency through design tokens.
Designed the public playground for IBM's Granite 4.0 models: the primary activation surface for the launch.
2.6× daily prompts · 3× unique users vs Granite 3.3
Designed IBM's human-in-the-loop coding agent IDE in one week: the proof point that secured executive sponsorship for SWE agent research.
Agentic IDE before Claude Code
Designed and built an AI-powered team knowledge base: drop a link and Gemini summarizes it and connects it into a live knowledge graph.
Design and development, solo
Designed an enterprise platform that turns AI risk assessment into a guided, automated workflow: readable by executives, rigorous enough for compliance.
Made AI risk legible to executives
Interaction experiments in the site's own material: the details that make AI interfaces feel alive.
Describe a vibe and a live agent restyles its own page, streaming its design decisions one token at a time, then auditing its own contrast and repairing what fails.
A fluid simulation with a terminal's voice. Drag, and energy is injected along your cursor's path, advected, diffused, decayed, then rendered as a ramp of monospace glyphs. No WebGL, no textures; just math and type.
AI product design and prototyping: Granite Playground, SWE agent IDE, AI risk platform. Client-facing demos for IBM Think.
Interactive visualizations of urban big data; published research.
Data-driven exhibition design with AIA New York.