🐷 piggy.ai — Know Where Your Money Went

Inspiration

It started with a group chat message that probably every college student has sent at some point: "wait, how is it only the 10th and I'm already broke"

Between the four of us, we had Uber Eats receipts we'd forgotten about, subscriptions we swore we cancelled, and a collective understanding that we were all pretty bad at this whole money thing. The scary part wasn't the spending — it was that we had no idea it was happening until it was already gone.

Every budgeting app we'd tried felt like it was built for someone with a salary, a mortgage, and a 401 (k). Nothing spoke to us. Nothing was honest with us. So we decided to build something that would be.

What We Built

Piggy is a personal finance analyzer built specifically for college students. Upload your bank statement as a CSV, and Piggy instantly breaks down your spending, projects your financial future, and delivers a personalised AI analysis in whatever tone hits hardest — a disappointed immigrant parent, an aggressive Wall Street bro, or a surprisingly wise Buddhist monk.

No signup. No bank connection. Your data never leaves your device.

How We Built It

Our stack: React + TypeScript + Tailwind CSS on the frontend, deployed on Vercel. Node + Express + MongoDB on the backend, deployed through Render. Gemini AI powering the financial analysis engine.

We did a lot of our early ideation and architecture planning in Claude, which helped us move fast in the first few hours when decisions matter most. From there, we split into parallel workstreams — one pair on the UI, one on the AI integration — and used GitHub to keep everything from catching fire.

For the AI layer, we built three distinct tone personas that take the same financial data and deliver completely different analyses. Getting Gemini to stay in character while still citing real dollar amounts from the data was trickier than expected — prompt engineering is genuinely an art.

We used Browser Use for web automation features, ElevenLabs for the voiceover in our demo video, and FigmaMake to rapidly prototype our UI before writing a single line of CSS. Clipchamp handled video editing for our submission.

The math behind the projection feature uses a simple forward model:

$$B(t) = B_0 + \dot{s} \cdot t$$

Where $B_0$ is current balance, $\dot{s}$ is the average daily net cashflow, and $t$ is days forward. Simple, but when that line trends toward zero on the chart it hits different.

Challenges

Honestly? Using this many tools simultaneously was chaotic in the best and worst way.

The hardest part was getting our AI pipeline stable. We had Gemini doing financial analysis, Browser Use running automations, and ElevenLabs generating audio — all at the same time, all with their own rate limits, response formats, and failure modes. There was a solid two hour window where everything was broken in a different way and we were all staring at each other.

We also went through probably a dozen different names and concepts before landing on Piggy. We almost built a WebReg auto-enrollment bot. We considered a grief tech app. We debated a debate simulator. The ideation phase was genuinely fun but also genuinely stressful when the clock is ticking.

The tone personas were a last-minute addition that we almost cut for time. We're really glad we didn't — it became the most memorable part of the demo.

What We Learned

That the gap between "this works" and "this feels good" is enormous and worth fighting for. We could've shipped something functional at hour 12. We kept going because we wanted something we were actually proud of.

We also learned that integrating multiple AI tools isn't just a technical challenge — it's a product challenge. Each tool has a personality, a set of constraints, and a learning curve. Getting them to play nicely together while building a coherent user experience is genuinely hard, and genuinely satisfying when it clicks.

And maybe most importantly, we learned we're all pretty bad at managing our money, and we now have a pig to tell us about it.

Built in 24 hours at DiamondHacks 2026. 🐷