Unsiloed AI: For an enterprise document platform
Priya Bhatt
I designed the interface for Unsiloed AI, an enterprise platform that turns messy documents (PDFs, scans, spreadsheets, slide decks) into clean structured data through four processors: Parse, Extract, Split, and Classify. The users are developers and technical teams in legal, finance, and healthcare, so the whole thing had to read like a tool, not a marketing site. I designed it in Figma and built the system to map directly onto shadcn/ui, Tailwind, and lucide icons, so the handoff to engineering was close to one to one.
I kept the login plain and rewrote every word so it sounds like a person, with one clear action and no filler.
The original document sits next to its structured output, and every line traces back to the exact spot it came from in the PDF, which matters when your users are lawyers and analysts.
A brand new account usually dead-ends on a blank screen, so I made the first thing you see point you toward your next step.
You build a schema field by field to tell the product exactly what to pull out, and citations stay one toggle away because traceability is the whole point.
I kept the file view fast and predictable on purpose, because people managing hundreds of documents want speed, not a reinvented browser.
Every processing run lands here with its status and timestamp, and the row menu only offers the actions that make sense for that job.
You ask a real question across everything you have uploaded, and each answer comes back with the source file and page number attached so you can check it.
Like this project
Posted Jun 7, 2026
Built Unsiloed AI's developer playground end to end: parse, extract, split, and classify flows on a cohesive zero-radius shadcn design system.
