Penny

• GIF

Inspiration

We have all had that moment at the end of the month where we look at our balance and think, wait… where did my money go? Most finance apps do a decent job tracking transactions, but they stop there. They show you exactly where your money went without helping you understand why it happened or how to change it.

I was inspired by the gap between financial literacy and real financial behavior. Knowing what you should do with money is easy. Actually doing it is the hard part. Penny was built to focus on the human behind the numbers by understanding their Financial DNA.

What it does

Penny is an AI-powered personal finance coach designed to predict and prevent invisible spending. Instead of sorting transactions into static charts, Penny looks at habits, routines, and behavioral triggers.

The Graph Penny visualizes spending habits as a node-based graph. It shows how things like late-night food orders, subscription overload, or weekend splurges connect to overall financial health. Your money becomes a map of your life, not a spreadsheet.

Behavioral DNA Penny generates behavioral insights like Impulse Sensitivity and Routine Stability. These scores explain why someone spends the way they do, helping them recognize patterns before they turn into problems.

AI Chat and OCR Users can snap a photo of a receipt and Penny instantly extracts the data. It then provides a gentle behavioral nudge, helping them decide whether that purchase actually aligns with their long-term goals.

How we built it

I built Penny using a modern, high-performance tech stack with a strong focus on visual polish, usability, and real AI integration.

Frontend The app is built with React and Tailwind CSS, using custom animations, blur effects, and a glassmorphic design to create a premium, tactile experience that feels more like a lifestyle product than a finance tool.

Visualization I designed a custom SVG-based network graph that dynamically maps spending habits as connected nodes. The graph updates in real time as behavioral signals change, making spending patterns feel organic and alive instead of rigid and tabular.

AI Logic Penny uses a real large language model to analyze transaction data, receipt text, and user behavior. OCR is used to extract structured data from receipt images, which is then passed into the LLM to generate behavioral insights, spending explanations, and personalized nudges. The model identifies recurring patterns, impulse triggers, and routine stability to produce coaching-style feedback that adapts over time instead of relying on static rules.

Challenges we ran into

One of the biggest challenges was making financial data feel approachable instead of intimidating. Designing a node-based graph that stayed readable while still feeling sophisticated required a lot of iteration and spatial UI experimentation.

Another major challenge was refining Penny’s voice. I wanted it to feel like a supportive, high-level coach, not a judgmental spreadsheet that nags users for every purchase.

What we learned

Building Penny reinforced that financial health is about 20 percent math and 80 percent psychology. Translating raw transaction data into meaningful, human insights was one of my biggest takeaways.

On the technical side, I significantly leveled up my skills in CSS animations and SVG path manipulation, especially when creating the smooth, flowing connections in the habit graph.

What's next for Penny

The next step is taking Penny from simulated to live. I plan to integrate real banking APIs like Plaid and leverage the full power of the Gemini API for deeper, more personalized coaching.

The long-term vision is real-time awareness. Imagine Penny noticing someone walk into a Starbucks and sending a quick nudge saying they have already had four coffees this week and skipping this one gets them five percent closer to a Japan trip. That is the future I am building toward with Penny.

Built With

• react
• tailwind
• typescript