App Orchid: scaling a design system behind an enterprise AI platform
Design work across App Orchid's product surfaces — dashboards, a natural-language "Ask" assistant, and the underlying component system that keeps both consistent in light and dark mode.
01 — Overview
Turning data into insight, in plain language
App Orchid is an enterprise AI platform built around a simple premise: people shouldn't need to write a query to understand their own data. The product spans vertical dashboards (like their Utilities Suite) and a conversational "Ask" assistant, both of which needed a design system robust enough to hold up across dozens of components, two color modes, and a fast-moving engineering team.
02 — Product surfaces
Dashboards that explain themselves
The Utilities Suite dashboard interprets a plain-language question back to the user as a set of filters — "Show all meters with material plastic, manufacturer is not Neptune" — before answering it, so people can trust and correct the system's read of their question. Below that, a plain-English summary sits above the charts themselves, translating the numbers before the user has to.
The "Ask" assistant extends the same idea into a full conversational interface — answering a question directly, then showing its work. Every answer surfaces its own caveats (excluded records, dropped nulls) inline, with the generated SQL available on request for anyone who wants to verify it.
03 — Design system
A component system built for handoff
With multiple product surfaces pulling from the same library, components needed to be specified once and trusted everywhere. Buttons and inputs were fully systemized — every variant, size, and state documented with usage guidance, so engineering could implement without guessing and design could extend without drift.
Buttons
Primary, Secondary, Tertiary, and Destructive variants, each in three sizes and four states (Default, Hover, Pressed, Disabled), with icon-before/icon-after configurations specified for every combination.
Inputs
Basic, Dropdown, and Multiline input boxes across every state (Active, Error, Warning, Success, Read Only, Disabled), plus compound patterns — Input Row, Input Table, Ghost Box, and Repeater — for dynamic forms.
04 — QA & collaboration
Bug triage, right on top of the design
Rather than logging issues in a separate tracker disconnected from the screens themselves, QA happened directly in FigJam — bugs and improvements pinned as color-coded stickies straight onto the flows, categorized by area (Text, Layout, Components) so the team could see exactly where problems clustered before a release.
05 — Testing across themes
Every component, twice
Because the product needed to work in both light and dark mode, every container and component in the system was tested side by side in both themes — not just spot-checked, but laid out as a full matrix so contrast, spacing, and state changes could be reviewed at a glance, with the team collaborating live in the file.
06 — Light and dark, end to end
Consistency from marketing site to product
The same theming discipline carried through to App Orchid's public site — documentation, blog, and forum all rebuilt to hold up in both light and dark mode, using the same color and elevation logic as the product itself.
07 — Client work & impact
Shipping across a growing client roster
The design system and workflow above weren't theoretical — they held up across real client engagements, each with its own scale and constraints.
Fidelity
Designed 25 custom apps and built the complete visual design and new feature UI for Contract AI, a separate Fidelity product, from the ground up. Consistently high-quality delivery led Fidelity to expand its scope of work with App Orchid.
GSEC & APM
Additional enterprise clients praised the work for making their product UI more consistent and modern, following the same systemized component approach.
Speed & process
Increased UI production speed to roughly half a day per design, and cut AI-assisted theme generation time from a full day down to about an hour. Built and maintained a Kanban board to track design system development across a 2-designer team.
QA at scale
Ran a structured UX QA audit — 75 issues segmented into UI, UX, spelling, and component categories — plus dashboard testing across 2 client engagements, surfacing 40+ issues on one dashboard and 20–30 on another before release.