Accuris · 2024–2025
Accuris Thread: uploading documents and managing requirement projects
90%TIME SAVED DEFINING REQUIREMENTS
90%DECOMPOSITION ACCURACY
Dishmatch: finding somewhere to sit down and eat
Overview
Dishmatch is a mobile app for finding food nearby. Some people want something quick and specific. Others want a long list to browse and narrow down. It handles both. I also used the project to try a new way of working: designing with Claude Code connected to Figma through the Figma MCP server.
Goals:
- Help people find places to sit down and eat, as well as delivery.
- Let users filter by cuisine, format, and occasion in one view.
- Test whether going back and forth between AI and Figma speeds up hi-fi prototyping.
ROLESole designer
TIMELINEMar 23 – Apr 1, 2026
RESPONSIBILITIESResearch, UX & UI design, prototyping
TOOLSFigma, Figma MCP server, Claude Code
The challenge
Most people search for food through a browser or a delivery app like DoorDash or Uber Eats. You can search for specific foods or cuisines, but the filters are built for ordering. They don't help much if you want a restaurant to actually go to.
format_quote
"But I want somewhere to sit down."
format_quote
"I want to choose between Mexican or Japanese food."
format_quote
"I have to search Google anytime I want something specific."
PROBLEM TO SOLVE
It's hard to find places to eat beyond takeout. Current apps don't let you filter for the kind of meal you want.
quick_reference_allResearch
Gathering insights
I looked at Google, DoorDash, and Uber Eats to see where they fall short for people who want more than takeout. Then I interviewed 10 people and turned the most common problems into personas and user journeys.
Competitive analysisarrow_forward
10 interviewsarrow_forward
Personasarrow_forward
Journey mapping
PERSONA 1
PERSONA 2
USER JOURNEY 1
USER JOURNEY 2
HOW MIGHT WE
How might we help people find the right kind of meal, quick or sit-down, now or later, in one place that's easy to get around?
lightbulb_2Ideation & design
Going back and forth between AI and Figma
I started by sketching problems and solutions on paper, which settled the core features, filters, and workflow. I turned those into structured prompts for Claude Code and added a custom rule so Claude could check reference images in a folder while refining its output.
01SketchesBrainstorming issues and solutions on paper.
02Claude Code mockupsPrompts produce a working base and prototype.
03Hi-fi in FigmaVia MCP: named components, local components, refined tokens.
04autorenewFeedback loopPush updates back to Claude, add features, re-import, refine.
SKETCHES
INITIAL PROMPT YIELD
FIGMA REFINEMENT
editor_choiceFinal design
The prototype
Robust filter suite
Set results to match what you're in the mood for. Want an Indian or Korean food truck nearby for lunch and a nice sit-down Japanese place for dinner? Dishmatch shows both on one screen.
View and save
Each restaurant page puts the ordering details first. Save a place to come back to later, and remove it when you're done.
USER FLOW MAP
monitoringImpact & next steps
Where it stands
Early feedback has been good. Next up is usability testing to see whether Dishmatch actually makes it easier to find somewhere to eat. I now use the same Figma MCP and Claude Code workflow for prototypes and redesigns as Staff UI/UX Designer at Democracy First.
10USER INTERVIEWS
9 daysCONCEPT TO HI-FI PROTOTYPE
3TOOLS IN ONE LOOP
boltFaster orderingHelp users go from discovery to order in fewer steps.
nutritionDietary filtersRestructure filters to include dietary restrictions.
collections_bookmarkSaved collectionsLet users group saved restaurants into collections.