AI products break most design patterns. Chat is not just a textbox. Agents are not just chat. Generative outputs need entirely new ways to review, edit, and trust what the system made. Most teams ship something that works in the demo and falls apart in real use.

I design AI product interfaces that hold up in the hands of actual users — chat experiences with memory and context, agent workflows with checkpoints and oversight, prompt-driven generation UIs, AI-assisted editors, and the boring-but-critical UX of error states, hallucination handling, and trust signals.

What I bring: real product design experience from shipping AI-heavy products (CRM with embedded AI, automation builders, AI-assisted mobile apps), familiarity with current model capabilities and limits, and patterns that don't just copy ChatGPT or Claude — patterns that fit your specific product and users.

Scope of a typical engagement: discovery on the AI feature or product, user flows for the AI interactions, screen designs in Figma covering happy path plus edge cases (loading, errors, low confidence, refusals, regenerations), prompt-input patterns, output review and editing UIs, and the surrounding product UI that makes the AI feel like part of a real product instead of bolted on.

Good fit: teams building an AI-first product, products adding AI features to existing flows, founders shipping an MVP with AI at the core, anyone whose AI feature works technically but feels wrong to use.