Finpulse

FinPulse — Intelligence behind every transaction.

Comment

Inspiration

Managing personal finances in India is messy at UPI payments, unclear SMS alerts, unknown merchant names, and zero context after a transaction. Most expense apps expect users to manually log everything, which breaks instantly in real life. We wanted to build something that understands transactions the way humans do, without adding friction.

What it does

FinPulse is an AI-powered personal finance assistant that automatically understands transactions, categorizes them, learns user behavior, and provides actionable insights. Using Gemini, it resolves unknown merchants, supports multimodal inputs (text, voice, receipts), tracks budgets and goals, and proactively alerts users when they deviate from disciplined spending—all through a conversational interface.

How we built it

We built FinPulse using a three-layer transaction ingestion system: screen reading, SMS & notification listeners, and the Account Aggregator (AA) framework. Gemini is used at the core for natural language understanding, merchant resolution, categorization learning, insight generation, and conversational reasoning. The system is designed to be local-first and real-time.

Challenges we ran into

Handling unstructured and ambiguous merchant names

Reducing notification fatigue while keeping tracking accurate

Designing AI explanations that are clear, not overwhelming

Mapping financial data into meaningful user insights

Accomplishments that we're proud of

Accurate auto-categorization that improves with use

India-first handling of UPI and SMS transactions

Conversational insights that explain why spending changes

A system that feels proactive, not reactive

What we learned

Good financial AI isn’t about more data—it’s about better context, timing, and trust. Users engage more when AI explains decisions and adapts to their habits instead of forcing manual effort.

Built With

Share this project:

Updates