Bono | Devpost

Inspiration

We are inspired by our personal experiences as international students and immigrants to create this web app that helps people with similar backgrounds navigate the US legal system.

What it does

No matter your background or comfortability with English, Bono acts as your guide to the US Legal system, giving you resources and asking you follow up questions that helps you understand the nuances of these laws in any situation. You press the green button and type in your prompt, and it will search for relevant laws that will help your case, then will help you narrow down what to do. Our promise is that no data is ever stored within the app, and we implemented a quiet mode to quickly leave the app in case of an emergency.

How we built it

Bono uses a small language model in Ollama to privately answer your legal questions without transacting your data with third parties. We used Jac Builder to create the prototype and the front end, and used its graph based logic to quickly send only relevant data throughout the multiple databases. We trained the AI model with RAG (Retrieval-Augmented Generation) on a dataset of common legal documents such as the US Constitution and USCIS Laws and Policy to answer your questions. Our approach was loosely derived from Recursive Language Models (arXiv:2512.24601 [cs.AI]) and the process of using previous prompts as the environment for further inquiry.

Challenges we ran into

One of our team members left during the middle of the competition due to an emergency situation that was out of our control, therefore we had to adapt and redistribute our workload, as well as scale back our approach. Our final product ended up being a lot more simple than expected.