screenshot_agent_mobile
screenshot_dashboard_mobile_1
screenshot_dashboard_mobile_2
screenshot_dashboard_mobile_3
screenshot_dashboard_mobile_4
screenshot_dashboard_mobile_5
screenshot_dashboard_mobile_6
screenshot_dashboard_mobile_8
screenshot_dashboard_mobile_7
screenshot_agent_desktop

Finrush — Voice-First AI Financial Assistant

Inspiration

Personal finance matters, but managing it often feels like a chore. Most tools require time, discipline, and lots of manual input—logging transactions, updating spreadsheets, categorizing expenses.

At some point, a simple question emerged: what if managing your finances was as easy as talking about them?

With the arrival of real-time AI agents like Gemini Live, that idea became possible. Instead of forcing people to sit down and organize everything manually, what if an AI assistant could listen, understand, and handle the boring part?

That's how Finrush was born—an attempt to put personal finance on autopilot, so people can simply speak (or show a receipt) and let an AI agent organize everything in the background.

What it does

Finrush is a voice-first AI financial assistant powered by Gemini Live.

Instead of manually entering transactions or navigating complex finance apps, users simply talk about their finances.

Examples of what users can say:

"I spent 45 euros on groceries"
"I invested 500 dollars in an ETF"
"How much did I spend this month?"

Finrush understands intent and automatically:

Logs transactions
Categorizes expenses
Tracks investments
Updates financial insights in a clean dashboard

In addition to voice, Finrush can analyze receipts, invoices, and tickets via camera. The system extracts relevant information and converts it into structured financial records automatically.

How we built it

Finrush combines a modern web stack with real-time AI capabilities.

Frontend Stack

Next.js + React 19 — Modern UI framework
Tailwind CSS v4 — Responsive styling
Browser-based camera capture — Image processing
WebSockets — Real-time communication

Backend Stack

FastAPI (Python) — High-performance API
Gemini Live API — Voice understanding & tool calling
Supabase PostgreSQL — Secure data storage

The Gemini agent converts natural language (and extracted image data) into structured financial operations: creating transactions, updating budgets, tracking investments.

Challenges we ran into

One important thing: I'm not a professional developer. I'm someone who loves technology and has always been fascinated by building things.

This meant building Finrush was also a personal learning journey. I had to learn many concepts in parallel:

Modern web frameworks
Real-time AI interaction
Building complex systems

Tools from the Google AI ecosystem were crucial. Using AI assistance (especially through Gemini) made it possible to move much faster than I imagined as a solo enthusiast.

Technical Challenges

Real-time voice streaming
Interpreting ambiguous natural language
Keeping financial data consistent across currencies and investment types

But the biggest challenge was simply learning fast enough to keep the project moving forward.

Accomplishments we're proud of

For me, the biggest accomplishment is simply not giving up.

Coming into this project without a professional software development background meant many parts initially felt almost impossible. But step by step—learning, experimenting, breaking things, fixing them again—I managed to bring the idea to life.

What Finrush represents

Finrush is technology that solves real problems:

Voice interaction for natural input
Financial tracking that actually works
An AI agent capable of organizing complex information

Most of all, I'm proud of having the determination to keep going until something that once seemed out of reach became real.

What we learned

Building Finrush reinforced several important lessons:

Voice is a natural interface for repetitive tasks like logging expenses
AI agents work best when natural language is combined with structured tools and clear data models
The current generation of AI tools is incredibly powerful — ecosystems like Google's AI platform are evolving so quickly that the future of personal assistants feels exciting

We are only at the beginning of what these technologies will enable.

What's next for Finrush

The next step is community feedback and testing.

As a solo developer, one major limitation has been the lack of large-scale testing. There are likely bugs, edge cases, and improvements that only real users will uncover.