FinRak

Inspiration

FinRak started from a real personal frustration: I was constantly juggling expenses across notes, screenshots, random apps, and half-updated spreadsheets. I wanted something simple, something that understood me when I typed naturally, like “Paid 1200 for internet last week” without forcing me into rigid forms or dropdowns.

Most finance apps felt either too basic or too bloated. I wanted a tool that felt conversational, intelligent, and genuinely helpful. That’s when the idea clicked:

What if expense tracking felt as easy as chatting with an AI?

What it does

FinRak is an AI-powered personal finance tracker that lets you manage your money through natural language. You can type something casually, “Spent $220 on clothes yesterday,” and FinRak automatically extracts the amount, category, date, and type of transaction.

It supports:

Income, expenses, pending payments, and pending receivables
Smart categorization and subcategories
A chat-like interface for quick entry
Google Sheets as a reliable backend
Deep analytics: trends, category breakdowns, insights, and spending patterns

In short, FinRak turns your everyday language into structured financial intelligence.

How we built it

I built FinRak using Streamlit for the UI, Gemini AI for natural-language parsing, and Google Sheets API for storage.

The workflow looked like this:

Designing a clean, responsive Streamlit interface
Integrating Gemini to interpret natural-language inputs
Building a structured transaction model with categories and subcategories
Connecting Google Sheets as a lightweight but powerful database
Creating analytics dashboards with charts, insights, and filters
Adding logic for pending transactions and due dates
Polishing the UX to feel conversational and intuitive

Every part of the system, from parsing to analytics, was built to feel seamless and human-friendly.

Challenges we ran into

I definitely hit some walls along the way:

Natural language parsing was trickier than expected. People phrase expenses in so many different ways.
Google Sheets API authentication required careful setup, especially with service accounts.
Date extraction from casual text (“last Friday”, “two days ago”) needed multiple iterations.
Analytics performance had to be optimized to avoid slowdowns with larger datasets.
Streamlit state management sometimes conflicted with the chat-like interface.

Each challenge forced me to rethink the architecture and refine the user experience.

Accomplishments that we're proud of

I’m proud that FinRak feels genuinely usable. Not just a demo, but something I could rely on daily.

Some highlights:

Building a natural-language interface that actually understands real-world inputs
Creating a clean, modern UI that works beautifully on both desktop and mobile
Implementing a full analytics suite with meaningful insights
Designing a flexible transaction model that supports multiple types and states
Making Google Sheets behave like a structured database

Seeing everything come together into a smooth, intelligent experience felt incredibly rewarding.