Luna | Devpost

Inspiration

We kept running into the same problem: people don't struggle to save or qualify for credit because they lack the money or the opportunity — they struggle because their banking app doesn't speak their language. Menus pile up, options overwhelm, and real financial opportunities slip right past them. So we set out to build something that talks with you instead of making you dig through screens, something that negotiates and adapts as you go.

What it does

Luna listens, converses, and builds. As you talk with it about credit and savings, it generates the interface in real time, right there in the conversation. There's no fixed form waiting to be filled out — the agent negotiates terms like amounts, timelines, and rates with you, weighing different factors about your situation as the conversation unfolds.

How we built it

We built the frontend in React Native, so the experience runs smoothly across devices. The backend runs on Python, where the agent lives: it interprets the conversation, decides what interface components to generate at each step, and drives the negotiation logic behind credit and savings decisions.

Challenges we ran into

Making it feel like a real-time conversation, not an app that loads things, was the hardest part. We had to generate dynamic UI from the backend while keeping the conversational flow tight, which meant rethinking, more than once, how we synced the agent's responses with what actually renders on screen.

Accomplishments that we're proud of

We're proud that we pulled off a real-time generated interface that feels fluid and natural — one that doesn't read as an app assembling screens on the fly, but as a conversation that simply takes visual shape as it goes.

What we learned

We learned how to generate dynamic UI straight from the backend in real time, syncing the agent's conversational logic with the interface it builds so the user never feels a gap between the two.