About Bite Roulette

Driving Question

How can technology help people discover authentic, diverse cuisines they would never find on their own?

The United States is one of the most diverse countries in the world. When many immigrants come to the United States, they share their unique cultures by opening restaurants. According to WifiTalents, in New York City, 60% of the restaurant workforce is immigrants. While these types of restaurants are usually not the most commercial or popular, they represent the beating heart of immigrant culture, where the most authentic food in the city is found.

Elevator Pitch

Bite Roulette is a restaurant discovery app that uses a spinning 3D globe to randomly land on a country, then finds nearby restaurants serving that country's cuisine using your location and the Google Places API. It turns the overwhelming question of "where should we eat?" into a fun, gamified experience that pushes people outside their comfort zone and toward authentic, diverse food they might never have tried otherwise.

Research & Beginnings
▼

My Story

The idea for Bite Roulette began when a friend invited me to his family's Georgian restaurant in Queens, a place with no Yelp presence, a handwritten menu, and maybe twelve tables. I'd never had Georgian food before. I left obsessed.

That meal made me realize how many restaurants like that exist in New York — incredible, authentic, family-run spots that never show up when you search "restaurants near me" because they don't have the budget to compete with chains for visibility.

Research

After defining the problem, the next step was figuring out how to actually build it. I researched different development platforms and first landed on Bubble.io, a no-code tool that lets you build web apps without writing backend code from scratch. I built an initial version of Bite Roulette on Bubble — it handled the database, UI, and workflow logic all in one place. But as the project grew more complex, I kept running into limitations: the globe integration was clunky, API handling was rigid, and customizing the frontend felt like fighting the platform at every step.

That's when a friend told me about Claude Code, an AI-powered coding assistant that lets you build fully functional web apps through natural language conversation. Switching to Claude Code was a turning point. It handled HTML, CSS, JavaScript, and API integration all in one workflow, and gave me complete control over every detail of the app. The flexibility was night and day compared to what I had before.

For the restaurant data, I researched available APIs and chose the Google Places API. It returns real-time restaurant data including names, ratings, photos, opening hours, and location; everything the app needed. Combining Claude Code's development workflow with Google Places API calls meant I could build a fully functional restaurant discovery app that pulls live data based on the user's location.

Future Goals
▼
Gain Userbase
I want to gain a consistent user base on the website
Gamify Further
Gamifying can make normal tasks much more fun and engaging
Make Website Profitable
All the costs for running the website and the API's make this website very costly. I will need to have a profit to at least make a net 0.
Market
Creating social medias could be a way to market my website to more people
Materials & Resources
▼

Materials

The app was built using Claude Code, an AI-powered coding assistant that generates and manages HTML, CSS, and JavaScript through conversational prompts. Restaurant data is pulled live through the Google Places API, which provides names, ratings, photos, addresses, and opening hours. Google Cloud Console was used to manage and restrict the API key. The Globe.GL library was used to display the spinnable globe.

Resources

Building the app required learning on the fly. Claude Code's ability to iteratively generate, debug, and refine code was the primary development method throughout the project. Google Places API documentation was used to understand available data fields and how to format requests correctly. Claude Code served as both the development environment and AI assistant, helping to troubleshoot logic, debug JavaScript, and work through API integration challenges.

Graphics
▼

Below are a few graphics of my project, showing a development diagram and the evolution of the globe from my first attempt to the current version.

Original globe built on Bubble.io
My first attempt at the globe, built on Bubble.io. Functional but limited in customization and visual polish.
New globe built with Claude Code
The newest version of the globe, built with Claude Code. Fully interactive with purple brand theming, twinkling stars, and smooth throw-to-spin physics.
Development process diagram
Development flowchart showing the build process from homepage creation through API integration and final polish.
Reflection
▼

This application process has been a learning experience. Going in, I had never used Claude Code, had never used an external API, and had never a clue about the depth of complexity in a simple website development. What I had initially thought was a simple process turned out to be a series of problems to be worked out one by one. Figuring out how to move information from the API to the frontend, getting JavaScript to function correctly with Claude Code's output, and understanding why the information I was getting back from Google was in the wrong format were all challenges I had to overcome.

The biggest challenge was the API. Getting the Google Places information to actually go into the application and display correctly was the longest part of the process. There were a lot of times when I wasted so much time figuring out the problem, only to realise that I was doing it completely wrong. There were a lot of times when I had to start over using a different method. While it was frustrating at times, I learned the most during those times.

If I were to redo this project, I think I would spend more time planning the data structure before starting anything. Many of the issues I ran into were because I made some early decisions that I would have to reverse later. Having a better idea of the data I needed, where it was coming from, and how it would flow through the application would have saved me a lot of time.