ReliefCal

Inspiration

We wanted to build an interactive database that gave users easy access to resources around them that they would otherwise not have entirely known about. This lined up really well with the GovOps prompt, so we decided to go after this category. We also wanted to make use of Cerebras' API to make our AI prompts to Google Maps' API a lot faster.

What it does

ReliefCal is a mobile app where users can find nearby disaster relief resources. These resources include police stations, fire stations, urgent care, shelters, and vendors that sell women products.

How we built it

We created an interface using Flutter and Dart where users can pick from specific categories for disaster relief resources. Picking a certain category will make a Llama query to a Google Maps API, which searches for nearby locations through a Google Map search given the user's location and need. Llama queries are processed through a Cerebras API, which provides fast AI inference. For queries that are not able to be directly resolved by Google Maps, we instead collect government data that includes resource addresses which are then searched on the Google Maps API. The results are then returned to the user in a list that provides detailed information and directions to available resources, which may include websites and/or contact information.

Challenges we ran into

This was the first time we were using an AI library that we were not used to, so our first few tests with the model gave us data hallucinations that were completely made up. We also ended up using all of our Cerebras credits at the end, so we weren't able to fully demo our final product.

Accomplishments that we're proud of

We were proud that we were able to get an app with a fully functional front-end and back-end, and using an AI model that gave us exceptional results on our queries. In our user research, we were also able to interact with a lot of fellow hackers, which provided us with a valuable experience of collecting interesting user data for our exploratory analysis.

What we learned

We learned how to effectively use the Cerebras API to make fast inferences for our product, and also how to collect data from various sources, such as from users or from open databases. This was also our first time working with the Google Maps API, which became one of the backbones of our app.

What's next for ReliefCal

Ideally, we would have more access to the Cerebras API so that our model is more accurate and there's less of a risk for data hallucinations. Additionally, we could provide access to more resources that users can choose from, such as additionally disaster housing options or available tree pruning services.

Built With

• dart
• flutter
• google-maps
• llama
• python