A way for you to document and share your home cooked meals and personal recipes. Easily keep track of your favorites, discover more for your specific goals, and document the details.
Role
Solo: product design, iOS build, strategy
Timeline
Four months
Platform
iOS · SwiftUI
Status
Built end to end, App Store submission next
At a glance
What it is
A social recipe app for people who cook. Document meals, save as a card, and share anywhere. Discover your ideal recipes here as well!
Social recipe app for people who cook.
My role
Solo, end to end. Research, design, brand, iOS engineering, testing, shipping.
Solo, end to end. Design, build, ship.
The Spark
My own problem. I love to cook and document my creations but I had no designated place to hold my recipes and photos.
My own problem. Nowhere to keep my recipes.
Built
A complete iOS app in SwiftUI and Firebase. Designed, coded, and tested solo.
Complete iOS app. SwiftUI and Firebase.
How I built it
Utilized AI workflows to translate my designs and product visions into an app.
I ideated and created a comprehensive work flow plan along with foundation design language for the app. AI acted as my engineering team to bring my ideas to a real, working app.
Claude Code, Codex & OrcaMy product engineering team
Base44, Loveable & PerplexityAccelerating iterations and research
FigmaDesigning and refining the screens
SwiftUIThe real native iOS app
FirebaseAuth, database, storage, cloud functions
My design and build process mapped out in Miro.
The spark
It started in my own kitchen.
Nommie started from something ordinary: dinner at my place. I cook most nights, I train,
and I want food that is genuinely nutritious and worth eating twice.
What I did not have was anywhere to keep any of it. My recipes lived in my head, in my
camera roll, and in a notes file I never reopened. People kept asking how I made things,
and every time I rebuilt it from memory.
The target user persona was just me at the end of the day. I then brainstormed the vibe and character of the app and got super excited about the potential.
Everything after this page came from that one problem.
The first page of Nommie. Hand-drawn: the concept, the first “Plated by” card, the warm cookbook identity.
The first vision
A personal cookbook that happens to be an app.
I frequently found recipes that I wanted to make and save on social media. The chefs would either put the recipe in their caption, personally DM it to you, or link out to a personal blog where they stored them. I thought: what if there was an app that acts as a personal cookbook and holds all of your recipes, but with a tangible “token” aspect that can be shared anywhere. So people who use social media to document their cooking can simply attach a recipe card with a QR code that links back to their Nommie account, and that becomes the consolidated place to store their recipes, and recipes in general.
The inspiration board: polaroid and cookbook warmth, hand-drawn line work, game-card structure.An early feed wireframe.
Type
Lora for headers, Nunito for body, Caveat for signatures.
Palette
Cream
Forest
Blush
Butter
Brown
I had a lot of fun figuring out how the token card would look. I wanted to make sure it would be interpreted as something similar to a Pokémon card, an artifact that could be posted or texted to friends and family.
The design thinking
A card built to travel.
The whole concept leans on behaviors people already have. The recipe card is the unit of value and the unit of marketing: made to be shared, replated, and recognized, the way a Spotify Wrapped card spreads. These artifacts shaped the design long before any talk of retention.
The recipe card is the unit of value and the unit of marketing. It is built to travel, with a QR code that leads back to the maker's Nommie cookbook. I wanted it to feel like a collectible artifact, similar to a Pokémon card.
After you create your first recipe, this flow appears automatically. I wanted the first card to explain what it is for, so the animation gives that purpose a clear moment.
Designing the recipe flow
Killing the friction in creating a recipe.
Once the first build was stable, I launched it to TestFlight and got a handful of beta testers using it in their own kitchens. I intentionally did not provide them with the premise to test if my idea was getting across or if I needed to adjust the delivery. A lot of what came back was about the micro-interactions, the small moments of friction that only surface when someone is actually cooking. One piece of feedback in particular pointed at something bigger.
“…I get tired after adding ingredients and then get lazy… if there was somehow a way to encourage filling the whole recipe card from conception, that would make it easier to share.”
TestFlight tester, first iteration
Before
Enter everything by hand
You typed a dish name, then added each ingredient and quantity one by one, then steps. The AI came in only at the end for macros. It felt like data entry, and people quit halfway.
After
Write it like you would tell a friend
Now you write your process in plain language, and one AI call reads it back as structured steps, gathers the ingredients, and, in the same request, infers the macros, cuisine, tools, and dietary flags as editable green suggestions.
Photo → Write → Generate steps and ingredients → Macros
Four decisions behind it
Structured step rows rather than a freeform paragraph the AI splits up. Your words are all still your own and AI helps to configure them into legible steps.
AI filled sections are lit up to indicate that they’re there and can be changed.
Quantities encouraged, not required. Precision is a choice, so you still get a card in about a minute.
Manual entry is always an option. AI is accelerating the process but never something you have to use.
Knowing when to stop
Knowing when to stop.
More design does not always mean better. I explored three ways to trigger a card export, trying to make the interaction more intuitive.
I was overcomplicating it. Evaluating all three made clear I was adding steps and inadvertently hurting retention.
Manual iteration hides this. When every version costs real effort, that effort clouds your judgment about whether it was needed at all.
AI removed the sunk cost. Accelerating the iterations let me judge them unbiased, and see the problem holistically.
Press and holdCleanest grid, hidden until you know.
Swipe upTidy, but a gesture to learn.
So yes, I ultimately decided to stick with my original simple export flow. The other flows overcomplicated that action and quietly forced the export. This realization helped me unearth another deeper, fundamental issue in my flow.
I came in attached to the recipe card and treated creating one as the main action. What testers helped me realize is that a share should never be forced. If I reduce friction everywhere else, making the card becomes easy, and whether someone shares it is up to them.
My dev takeaway, after the first round of testing
Stepping back
I had to stop designing screens and start designing a reason to return.
I am a design-forward person. I love the interactions and the visual appeal but I also love product. Testing forced me to zoom all the way back out again. A cute app with no reason to come back is not a good product. This was the moment I grew the most in this journey. I learned to carry the business perspective, retention and growth, and not just the screen in front of me.
Finding focus
Finding focus.
People enjoyed the look of the app but retention dropped off quickly because there was no real reason to come back. Part of that was because not everyone cooks everyday, part of that was there weren’t a ton of users, but the biggest part was it lacked context and discovery.
The app was static.Nothing changed between visits, so there was no pull to open it again.
A cold-start content problem.A new cook landed on an empty, generic feed.
The identity was too broad.Nommie needed a stronger, clearer identity instead of trying to be a recipe app for everyone.
The fix. Narrow to one high-intent persona (the active, health conscious cook), and add an engine that makes the app feel tailored from the beginning.
My miro flow mapped out that helped me recognize where action ceased and helped me figure out what was needed.
The research
Everyone tracks. Nobody plates.
To commit to the wellness direction, I researched other apps in the category.
of Americans tried to eat more protein in 2024. Most discover food on social. A rising, reachable market.
The competitive teardown
App
Biggest strength
The opening for Nommie
MyFitnessPalThe default
Unmatched database breadth.
Bloated, ad-heavy, impersonal, and lonely.
MacroFactorThe coach
Best adaptive coaching algorithm.
Pay-to-play, solo, tuned for serious dieters. No community.
CronometerThe lab
Nutrient depth and accuracy.
Clinical and joyless. No cooking, no people.
LifesumThe pretty one
Best-in-class visual design.
Gorgeous but still solo. No real feed or cooking.
Cal AIThe shortcut
Lowest-friction photo logging.
Estimates only. No recipes, no community.
What I built
A profile that knows your cook, and a feed that uses it.
The payoff is one system: an onboarding that learns what kind of cook you are and what your nutritional goals are, which then feeds a recommender that gives personalized recommendations.
01 · Know Your Cook onboarding
Teach the app your taste before you do anything.
A warm onboarding that explains the app’s premise, and then goes into asking about cuisines you enjoy, your nutritional goals (very soft), kitchen tools you possess, what you can’t eat, and how long you typically spend cooking. It then writes a TasteProfile. Since the app is not a rigid health app but still has a health forward philosophy, it briefly explains the benefits of different types of macros on the body to subconsciously remind the user of that.
02 · The adaptive feed
A real recommender, no ML infrastructure.
Home has a Friends / Explore toggle. Explore ranks every recipe with a content-based scorer (DiscoveryRanker): cuisine affinity, diet overlap, protein-fit weighted by how much you care, time budget, a goal nudge, log-capped popularity, and freshness decay. Allergens are a hard filter; missing tools soft-penalize in Explore and hard-filter in Home. Jitter and cuisine round-robin keep it fresh, and it falls back to popularity for brand-new cooks. Collaborative filtering slots into the same function later.
03 · The balance mirror
A weekly readout, not a calorie tracker.
On your profile, there is a mini dashboard showing your goal’s recommended macro split against the average of what you actually cooked that week. It is meant to show a soft holistic overview of how you typically cook and what macros you could try to incorporate more of. Never any rigid language or diet culture present.
A line I would not cross
The nutrition layer is a balance mirror, never a calorie tracker or a diet-shaming tool. It reflects how you typically cook back to you, warmly, and stops there. Ethical wellness and a home cooking encouragement focus is the goal.
04 · My Kitchen
Room for a little fun.
The kitchen tools section in settings I had a lot of fun making. Settings always feels so stale and I thought, if there is an option to select what tools you own, wouldn’t it be fun to display it in a way that has it sit on a kitchen counter? It still works properly as well and I hoped it would bring a bit of whimsical energy to this normally boring section.
A personal reflection
What I am taking with me.
I designed and built Nommie end to end, and the process changed how I think about making a product. Using AI as my engineering team gave me room to test ideas quickly, while TestFlight feedback taught me that a share should never be forced. The biggest growth came from learning to think beyond the screen and design for a reason to return.
The app is built end to end. App Store submission is next.
Nommie Q&A
01
Is this a diet app?
No, and I was careful about that. No daily target, and nothing turns red. The nutrition layer is a weekly mirror of how you have been cooking, and then it stops talking.
No. A weekly mirror, never a daily target.
02
Do I have to weigh everything?
No. Quantities are encouraged, never required. Precision is a choice you make per recipe, which is why a card still takes about a minute.
No. Quantities are encouraged, never required.
03
Does the AI write my recipes for me?
It never invents content. You write the process the way you would tell a friend, and one call reads it back as structured steps and does the macro math. Anything AI filled glows green until you touch it, so you always know what is yours.
No. It structures your own words, nothing more.
04
What if I do not own a stand mixer?
Then you will not be shown recipes that need one. You build your kitchen in settings, and that feeds the recommender. Missing tools soften a recipe in Explore and filter it out at home.
Then you will not be shown recipes that need one.
05
Why a card, and not just a post?
Because a card can leave. It carries a QR code back to your cookbook, so a recipe shared anywhere else still points home. The card is the product and the marketing at once.
A card can leave. Its QR code points back home.
06
What if I want to keep my recipes to myself?
Supported, by design. Testing turned up two kinds of cook: one shares to be seen, one just wants their wins in one place. Sharing is offered, never forced.
Supported by design. Sharing is offered, never forced.
07
Did you really build the whole thing?
Yes. Research, brand, illustration, the SwiftUI app, and the Firebase backend, solo. AI was my engineering team, working from specs and a design system I wrote.
Yes. Research, brand, the SwiftUI app, the backend, solo.
08
Can I download it?
Not yet. It is built end to end and has been through TestFlight with real cooks in their own kitchens. App Store submission is the next step.
Not yet. Built end to end, App Store submission next.
From real kitchens
Made by people who cook.
A closing slideshow of real recipe cards plated by Nommie users, back to the human feeling the whole thing started from.
