MealFinder

Inspiration

There have been many times where I, as a young college student surviving on chump change, have Googled 'cheap vegetarian food nearby' so I could enjoy an after-school meal with my friends who can't eat meat. More often than not, the search results would prove less than helpful; restaurants that were vegetarian were out of my meagre budget, and budget-friendly options rarely had vegetarian options. Even if they did, I would have to hunt for them myself on the restaurant's website by downloading their menu; the information would hardly be available to me at a glance. Worse still, even if there were such eateries, they would be in a completely different part of a city! This has been the source of much hunger-fuelled frustration, and I know I am not alone in this. Schools are where we meet and bond with people from a myriad of different backgrounds (especially in a cultural melting pot like Singapore, where I am from), and food is something that should universally connect us - we should all be able to spend time together and have fun over a scrumptious meal, without having to worry about burning a hole in our already shallow wallets or being able to find a place which accommodates to all of our dietary needs.

What it does

MealFinder uses generative AI to recommend eateries that are nearby to a specific location. Any dietary requirements and cuisine preferences can be specified, and MealFinder displays any eateries that are nearby, along with popular menu items that fall within a specified budget.

How we built it

MealFinder was built using AWS' PartyRock platform. I built the app using the platform's app generator and refined the prompts myself.

Challenges we ran into

It was hard at first to pinpoint what improvements to make to the prompts as I was not too familiar with what good prompt engineering principles were. After reading more about it, I made necessary adjustments to make the app behave how I wanted it to.

Accomplishments that we're proud of

I am glad to have made an app that I myself can use on days where I will be in school and even outside of school if I happen to be in an area of the city that I am unfamiliar with. I believe frugality is an important virtue for everyone, but particularly so for those who stay in cities with a high cost of living like I do. MealFinder is a tool that I believe can help me, and indeed many others, to make good financial decisions vis-à-vis the meals we consume on the go.

What we learned

I learnt the basic principles behind what makes a good prompt, as well as the essence behind foundational models and how they can be extremely helpful for niche use cases.

What's next for MealFinder

MealFinder is admittedly not perfect, and may be constrained with the current state of foundational models. As generative AI continues to develop out of its nascent stages, I hope the accuracy of MealFinder's recommendations will continue to improve at a commensurate rate to better serve people across the globe, no matter where we are (and how empty our wallets are!)

Built With

• partyrock