CalCommute | Devpost

Inspiration

Skye and I (Mariana) have been BART users as Bay Area locals, and have both found fascination in how AI and LLMs can help us navigate our trips as busy students. Even though we are community college students, we know how crucial public transportation is, specially when we transfer to a 4-year. Since we are at UC Berkeley, we decided to make it local to Cal students, however this idea can be extended to any campus. We envisioned an app that is much more personalized to our needs as students, with a focus in making Commute Buddies, sharing rides and meeting people with similar commutes. Additionally, similar to how it would be to ask a local, we integrated AskOski, a chatbot which can also access live bart and public transport data, planning out students' own trips in a personalized way. The chatbot will make recommendations to better optimize the students commutes, making use of their profile to connect them to potential commuter buddies sharing their trip.

What it does

The app has 4 subsections. Planner: the student can easily type their origin and destination, or select from saved places in their profile such as home, work, or school. The app will then release suggested routes including live public transport data. AskOski: Interactive chatbot acting as a berkeley local to help users navigate their specific trips. Ask Oski can display a map of the user's route and also include live data of public transit. Buddies (BETA): Find people with similar commutes to make CommuteBuddies to travel with, or even carpool and uber together. Profile: This is what makes CalCommute even more personal. If you choose to do so, you can include your hometown, hobbies, and common destinations. Any of these can be made private. You can interact with other users and get to know other Cal people that might be traveling to the airport, you can split an uber with them, and get to know other members of the Cal community while helping the environment and saving some money!

How we built it

As complete beginners, we were initially unfamiliar with how to incorporate LLM’s or just AI in general to our very general idea that involved accessible public transportation information in college. However, after attending Google’s gemini workshop we found a tool that allowed us to untangle and see some structure in our idea. This is how our app started. With a very basic design, we started adding more prompts and functionalities of the idea we envisioned. First, we wanted a profile section to include information about each student with the idea of establishing CommuterBuddies. Then, we wanted to further develop our basic chatbot option so that it can include a link to the google map that shows the trip that the user solicited, centralizing bart data and google maps with the chatbot. We accessed google maps and bart API keys to incorporate their data into our app, a task we were intimidated by, but found not to be bad at all! AIStudio allowed us to make our suggestions and edits in English, but we also learned the functionalities of each file in our project, making more specific edits for java in terms of content, html for format and css for design.

Challenges we ran into:

Sometimes AI studios wouldn’t listen to our queries and would respond as if it took them but our app wouldn't show those changes we solicited. We realized that by asking more specific questions to each file and their functionality, AIStudio would better respond to those queries. Initially, we weren't able to make the app work with the live data, and the app was just generating mock data that was inaccurate. This is when we realized that that was what API keys were for! We accessed the bart API and google maps API keys successfully incorporating it in our code, and now through the app you could access the commutes either from the chatbot option or the planner.

Accomplishments that we're proud of

We are very proud of figuring out how to incorporate the live data ACCURATELY and also being able to integrate a great tool like google maps into our app. We are also very happy with how the user interface turned out, since we feel like it's very intuitive and useful. To further how useful this app would be, we were able to help one of our mentors find her way home to salesforce park throughout our app!

What we learned

There are AI tools such as AI studio that can allow the creative process to be much less intimidating for beginners. Even though the first app version created by AI studio was very basic and was making up information, we slowly figured out how to include the real life data. We also learned about the functionalities of each separate type of file, and how we could prompt AI studio to change to our own needs. We finally learned what an API key is and why they matter!

What's next for CalCommute

CalCommute Buddies portion is still based on mock data since it hasn't been released to the public and we don't have any users, so it would be cool to implement that. We also have to consider privacy concerns when sharing location and commute data in social media, and what the best way to implement CommuteBuddies in a safe way.