2026
• Founding AI Design Engineer
Overview
Timeline
2026 · MVP V1 live
Team
Minzhe Emily · AI design engineer
Eng. partner · Backend & models
Background
Every new shoe is a costly guess
Violet St is a fashion house that has worked with Takashi Murakami, Kanye West, Wang Yibo and Fan Zhendong. Every new shoe is a bet: 6+ months of development, $100k+ in samples across every size, and if the shoe does not sell after mass production, $1M+ gone. Our CEO asked a simple question: can we know earlier, before we pay for all of that?
Ask everyone first, then go deeper
A real focus group asks everyone first, then follows up with whoever said something interesting. So the group interview and the one-on-one became one flow, and no context gets lost in between.
A score means nothing without a reference
Different AI models score the same shoe differently, so one number alone means nothing. Every test runs against a shoe already on sale: same people, one thing changed, and we read the gap.
The report answers marketing's questions in order
I kept asking marketing and ops what they need to know, and the report follows that order: who won, three findings with evidence, what to do next, then competitors and audience. Raw answers come last.
AI Director: ask why, right next to the data
People want to ask why. Instead of pasting a screenshot into ChatGPT, they ask the AI inside the result page, with all the data behind it.
2,000 people
2,000 AI shoppers built from real customer data, each with about 500 details about who they are and how they buy.
Their own feeds
Before answering, each shopper scrolls an Instagram-style feed of real brands and competitors, so they react the way a real customer would.
Three models
Three AI models answer every question. We keep the middle answer, so one model's quirks never decide the result.
After a shoe launches, its real sales go back into the system and get compared with what the AI predicted. Every launch makes the next prediction more accurate. Up next, in order: ad testing, store selection, and a virtual shelf.
In summary
From one sentence from our CEO to a live MVP: PRD, user flows, design system and prototypes, built side by side with one engineer. Three things I would bring to the next team:
Zero to one
Turned one sentence from the CEO into a PRD, a user flow and a live MVP. Chose the AI market over 3D renders, trend tools or an agency.
Product thinking
Merged two features into one interview, compared every score against a shoe already on sale, and ordered the report around what marketing asks first.
