A design system that scales a bottler platform — and an AI workflow that speeds up how it's built

Design and engineering had no shared source of truth. Components drifted between what was designed and what shipped, patterns were re-litigated on every feature, and discovery-to-design ran slowly and sequentially — roughly three weeks of research, then three of design.

The platform had to scale across web, chatbot / mobile web, and native Android and iOS apps without every squad reinventing basic UI.

I inventoried 60+ existing Figma files to centralize the patterns and components that were scattered across them, then built one clean library of roughly 50 components covering the platform's interfaces. Foundations — spacing, typography, sizing, motion — were documented as a variables/tokens architecture applied directly to the components: an abstraction layer that keeps the system scalable and consistent for a platform with 100k+ monthly recurring users, and one that spans web, chatbot / mobile web, and native Android and iOS.

The bigger question was workflow: how to reduce discovery time from six weeks to two or three. I worked on agent architecture and on finding the most efficient flows — automated research agents doing market study, benchmarking, and data analysis; AI tools for rapid prototyping and validation; and a Figma-to-specs agent that cut manual hi-fi work from a week to about a day. The handoff is built to feed a downstream pipeline of development agents with the full discovery context already attached.

I led a shifting team of varying seniority, aligning expectations with product and client and giving designers ownership — coaching them on design systems so they could help maintain it too, rather than depending on a single author.