Burpy closed in 2025. Personal names, phone numbers and partner logos on screens are replaced.
Problem
A diet plan nobody follows
Most people we talked to had the same story. The doctor gave them a list of foods to avoid, and nobody told them what to cook on Tuesday. Dietitians are very expensive in India, and calorie apps had no idea what to do with a plate of dal (lentils) and rotis (flatbreads).
The founder’s own research since 2021 found a second problem: paid diet plans are templates people drop.
One adaptive plan had to fit three kinds of eater:
people managing diabetes or high blood pressure
students and workers on PG (paying-guest) food, tiffins (delivered home-style meals) or hostel mess (shared dining hall) meals
families where a cook or house help makes the food
The condition manager
- Who
- People managing diabetes or high blood pressure
- A plan must
- Set limits after a diagnosis, or focus on prevention for a risk
- What it changed in onboarding
- Each condition asks: diagnosed, or at risk?
The PG or hostel eater
- Who
- Students and workers on PG (paying-guest) food, tiffins or hostel mess meals
- A plan must
- Work without a kitchen, around a fixed menu
- What it changed in onboarding
- From outside lists meal services like tiffins and hostel mess
The family with a cook
- Who
- Homes where a cook or house help makes the food
- A plan must
- Fit food someone else prepares
- What it changed in onboarding
- At home asks who cooks, for how many, and whether a child eats too
Three eating situations, one plan. Reconstructed for this case study from what we knew.
How it started
The recipe app changed its mind
Burpy began as the founder’s recipe app. In January 2024, I tore down six food apps. They were thin on Indian home food, ignored health conditions and never adjusted when you missed a meal. In design reviews, we moved Burpy to a meal plan.
- Dec 2023Joined as design intern
- Jan–Feb 2024Core moved to meal planning
- Jun 2024Full-time; private beta
- Apr 2025Burpy closed
Change 01Built in beta
Hostels don’t have kitchens
A plan that assumes a kitchen fails a hostel student on day one, and none of the six apps asked. So onboarding asks where meals come from, and every answer feeds the AI’s first plan.
From outside lists meal services like tiffins and hostel mess. At home asks who cooks, for how many, and whether a child eats too.
After the beta, I added a follow-up per condition: diagnosed, or at risk? A diagnosis needs limits; a risk, prevention.
Meal source
Where do you get most of your food?
Meal service: tiffin service, PG food, canteen, hostel mess
Which meals do you rely on? Breakfast, lunch, snacks, dinner
For whom is the food prepared? Family, or myself only
Who prepares your food? Me, a family member, a cook, house help
Total number of members
Is there a kid in the family? (12 years or below)
AI plan generation uses these answers
Health conditions
At risk of, or already have: diabetes, high blood pressure, heart conditions, kidney disorders, thyroid…
For each condition: diagnosed, or at risk?
The plan sets limits
The plan focuses on prevention
- Edge caseMeal service is optional in v1, so people can skip it.
- Trade-offEach person sees one path; the file holds both.
- VersionThe diagnosed-or-at-risk follow-up came with the revamp, built after the beta.
One answer routes the rest of onboarding. Reconstructed for this case study from the screens.
Change 02Designed, not built
Swap dishes, see the week change
The AI’s first plan is a guess, and none of the six apps let a plan bend. With a locked plan, one disliked dish meant regenerating the whole week.
In Customise your plan, people pin dishes and Regenerate rebuilds around them. A top card shows the week’s calories and nutrients changing live.
The budget is weekly because Indian weeks are uneven: a Tuesday fast or a Saturday wedding shouldn’t fail the week.
In detail: what Customise your plan lets people do
- Pin or unpin a dish; Regenerate keeps pinned dishes and rebuilds around them.
- Drag a meal to another slot, or copy it to other days, with a choice to replace what’s there.
- Add food from search, recommendations or favourites, with an undo after each add.
- Read insights under the budget, such as low vitamin C, each with an Optimise action.
- Save the plan with a tag (high protein, detox, balanced) and a duration, and generate a shopping list from it.
Change 03Built in beta
Snap it, fix it in words
People we talked to had given up logging home food in other apps: too much typing, too much guessing. In Burpy, you snap a photo and the AI lists each item’s calories.
When the AI is wrong, and sometimes it will be, a serving ruler fixes the amount. Edit with AI fixes the items in plain words: type “Add one cup of yogurt” and the meal updates.
The whole app
Twelve areas, all at once
Burpy grew to about 500 Android screens in twelve areas, and a peer designer polished trackers and paywalls. That’s where the trouble started.
In detail: every area in one line
- Onboarding: goals; diet in India’s own terms (Jain, meaning no onion, garlic or root vegetables; eggetarian, meaning vegetarian plus eggs; non-veg days, when a family eats meat); health; regional cuisine; meal source; cooking skill; about you.
- Home and discovery: a personal feed and Cook With What You Have.
- Search: recipes filtered by diet, time, meal type, ingredient and technique.
- Recipes: my redesign of the founder’s first screens: steps, nutrition and a health score, with tappable ingredient pages.
- Meal planner: the AI plan (in beta as a plan view); Customise your plan and saved plans (designed, not built).
- Logging: Snap and scan a meal, Edit with AI, and a Burpy AI call that nudges you to log a missed meal.
- Trackers and analytics: nine trackers, among them weight, water, sleep, period and blood pressure, plus condition signals from food already logged (carbs for diabetes; sodium and potassium for blood pressure).
- Social: feed, messages and support groups for people with the same goal (designed, not built).
- Groceries: a list from the plan, and store comparison (designed, not built; checkout later scrapped).
- Create: posts, articles and recipes, with AI import from text or a link (designed, not built).
- Family and cook: invite family members (unfinished), and share the plan with a cook.
- Subscription: free trial, plans and paywalls.
What failed
We became the template
We set out to replace one-size-fits-all diet plans. Then we designed a one-size-fits-all app.
I lost the two arguments that mattered most. First, I wanted a shorter onboarding with a skip option. The founder saw onboarding as our USP, and the AI kept asking for more data. Beta testers called the 14-frame version long, and the revamp made it 21.
Second, I argued for launching the planner alone. He wanted the full app first, and features kept arriving. Customise your plan, the planner’s heart, was still unbuilt when I left.
We decided to think about navigation at the end, and the end never came.
Craft
Making a long onboarding finishable
If onboarding had to be long, it had to feel short. A step counter and a cheerful mascot carry people through.
While the AI builds the plan, a swipe game collects food likes that shape it.
Few people we talked to counted calories, but they do know rotis. So Adjust goal adds portions to the numbers: 3 rotis a day becomes 1.
Outcome
A small beta, a quiet ending
In June 2024, a private beta opened to fewer than 30 friends and family. 8 of the 12 app areas reached them, including onboarding, the AI meal plan and Snap logging. Burpy closed in April 2025, after I left.
With a real launch, I’d measure onboarding drop-off and first-plan edits.
“I hired Amandeep on trust as a fresher [with no work experience], and he never broke that trust. He started stronger in UX than in UI and caught up on UI over time. His arguments for a better user experience were very good.”
Names left out for privacy. Happy to share references.
Retrospective
Ship one loop first
Next time, I’d put the planner alone in front of five testers, on structured calls, before designing a second area. Their answers would settle scope better than my opinion did.