goHeather
AI contract review for businesses that move fast.
I joined an existing product and rebuilt it. A new design system first, then every surface migrated onto it, and new product on top: the AI chat that edits contracts, workflows, playbooks. Design and features at the same time, shipped in small pieces while the app stayed live for customers.
What I walked into
The contract review engine worked. Clause detection, risk scoring, the hard part was real. What surrounded it was a table of past reviews and no way to actually talk to the thing, so the product read as far less capable than it was. My first deliverable was not a ticket, it was an audit: 59 findings across seven axes, four of them blockers.
Design system before features
Migrating screen by screen would have produced a hybrid app that looked broken in the seams. So the system came first: tokens, type scale, spacing, and a component library of 59 components that every later surface could be built from. It is the reason the rest of the work went fast.
Then every surface, in small pieces
Home, contract review, playbooks, the editor: each one migrated onto the system as its own stacked pull request, reviewed and merged while the product stayed live for customers. The AI chat, the workflows and the connected sources were not moved over, they were built. 464 pull requests landed. Nothing shipped as a big-bang rewrite because nothing had to.
Product work, not just execution
The title says AI Product and Design Engineer and all three words did work. I scoped the tickets, wrote the product specs, designed the flows in Figma, and then shipped them myself.
Stack
TypeScript · React · Vite · Tailwind CSS · TanStack Query · Zod
Supabase · Hono · pnpm monorepo · Figma · Linear · Graphite stacked PRs