Rallie | Devpost

Inspiration

I am a student activist, and I have faced bureaucratic systems often. My team and I are building the tools we wish we had when we were learning about new systems, using advanced AI technologies

What it does

We have a dataset we collected through interviewing UC Santa Cruz (UCSC) students at the hackathon using an agentic interviewer. You, as a student of UCSC, can learn from the trials of your predecessors.

How we built it

We built it using Cloudflare for our infra, Gemini for our inference, MongoDB for our database, FastAPI for our backend, and a React front end.

Other tools used: MongoDB, BAML, LangChain, Pydantic, Vite, Typescript, Python.

Challenges we ran into

A challenge I spent a ton of time on was configuring a cloudflare tunnel. I had to step into unfamiliar territory, and failing meant letting my team mates down. Something I found difficult was learning how to communicate over http from the back end to the front end. It required a lot of coordination and exposed some oversights on both ends that we would have otherwise overlooked.

Accomplishments that we're proud of

Getting the front end and back end connected
Having time to gather data from real interviews
Getting the cloudflare site working

What we learned

Database frameworks
Different perspectives on bureaucracy

Rallie's future:

Pre Release

Interview UCSC students to gather pre-release dataset
Perfect RAG pipeline
Add Wiki feature
- Chat history feature

Beta Testing

UX Testing, nail interaction design for agents
Focus on gathering early adopters and building community

Release

A final version of Rallie is one where we have a stable knowledge-base of UCSC, that will last for decades.
We aim to expand to other schools and communities, because the problems we face here at UCSC are not unique.