They asked me to redesign a meal page. I built the machine underneath it.

How it started

The service worked. The product had to catch up.

Livin’s in-person experience already worked: chefs cooked in members’ homes and people loved it. The digital product had not caught up. The founder came from marketing, one engineer held the codebase, and nobody owned product. I joined in Q1 2026 to take that seat, brought a product designer from my studio, and started by looking at everything a customer touched.

Q1 to Q2

The menu PRD became the platform decision

In Q1, I wrote the menu PRD around both sides of the problem: what customers needed to find and trust a dish, and the operational work required to keep the catalog current. When I presented it, the next question was how the team would actually run it. I proposed moving the catalog into Sanity and building the marketing site on top of the same headless system.

The menu

630 dishes, modeled once, reused everywhere

It started with a basic customer question: what are the macros? Livin was already describing dishes as metabolic-health or GLP-1 friendly, while customers still had to message the team for nutrition information. I set out to put the macros on the menu. That became the work of giving 630 chef-written recipes a nutritional foundation we could stand behind.

Demand

The onboarding was a system, not a quiz

A relevant landing page only works if the next step remembers why someone came. I built the onboarding system in Sanity so a person arriving for fertility, metabolic health, or fitness could move through an experience shaped around that need, instead of starting over in a generic form.

Supply

Keeping the menu fresh was the real bottleneck

As the catalog grew, keeping it fresh became a long manual process. The team needed to research new ideas, understand what the current menu was missing, create dishes, and review them before anything went live. I built a menu builder that brought that work into one system.

doesn't ship

Brief

plain language

Generate

recipe + nutrition + copy

Verify

checked against the goal

Publish

chef-ready

“Marathon training, mostly plant-based, kid only eats mild food.”

Charred Lamb Kebabs

Lamb kebabs · Whipped hummus · Cucumber salad & rice

44g proteinGluten-freeHigh-protein

High-protein

Charred Lamb Kebabs

Whipped hummus, rice & cucumber salad.

“Marathon training, mostly plant-based, kid only eats mild food.”

Charred Lamb Kebabs

Lamb kebabs · Whipped hummus · Cucumber salad & rice

44g proteinGluten-freeHigh-protein

High-protein

Charred Lamb Kebabs

Whipped hummus, rice & cucumber salad.

Expand

The foundation

I built each surface where it belonged

By Q2, I was shipping Livin's external menu in Next.js. The existing-customer menu needed to stay in Rails, so I built a working HTML prototype with the full add-to-cart flow. That gave the Rails developer something concrete to implement and kicked off the rest of the product work.

Design to production, in one pass

←01 / 04→

150+ tokens, one source of truth.

Later, when we built the customer AI chat, we kept that surface in Next.js. I handled the interface and the behavior: it carried customer context, pulled relevant menu items, and could take a custom order. The referral flow used the same product language to turn a member’s share into something personal.

Measurement

The homepage number was not a homepage number

Every week, the same question came up: how's the homepage doing? I had been building and reshaping dashboards in PostHog since the beginning, trying to get the team a read we could trust. I could see conversion move, but I could not tell whether the page had changed or the audience had. When I dug in, the answer was channel mix. A cheap channel could bring in more people and lower the blended rate while the homepage behaved exactly the same.

In numbers

Numbers as of July 2026.