SKIP TO CONTENT
FINSIGHT cover
MOBILE APPLICATION / ARTIFICIAL INTELLIGENCECurrently runs locally · Version 2 planned for public release

FINSIGHT

FinSight — AI-Powered Financial Intelligence Application

01

OVERVIEW

FinSight is an AI-powered financial intelligence mobile application built using Flutter. It processes bank transaction SMS messages, converts them into structured financial information, categorizes expenses, monitors budgets and provides financial analytics.

02

PROBLEM STATEMENT

Most users have no clear picture of where their money goes after UPI payments, and typical expense trackers depend on manual entry.

03

OBJECTIVES

  • ▸ Process bank transaction SMS messages automatically.
  • ▸ Convert messages into structured financial information.
  • ▸ Categorize expenses and monitor budgets.
  • ▸ Provide financial analytics.
04

PROPOSED SOLUTION

An app that captures UPI transactions from bank SMS messages automatically, parses them with a custom NLP engine, stores them locally, categorizes expenses with machine learning, monitors budgets in real time and predicts the probability of UPI transaction failure.

05

FEATURES

  • ▸ Automatic UPI transaction SMS processing
  • ▸ Transaction extraction
  • ▸ NLP-based transaction processing
  • ▸ Automatic expense categorization
  • ▸ Budget monitoring
  • ▸ Budget alerts
  • ▸ UPI transaction failure prediction
  • ▸ Financial analytics
  • ▸ Income and expense visualization
  • ▸ Offline-first functionality
  • ▸ AI-powered financial chat interface
06

TECHNOLOGY STACK

FlutterDartRiverpodHivefl_chartCustom NLP EngineXGBoostscikit-learnFastAPI
07

SYSTEM ARCHITECTURE

Layered design: SMS collection → NLP processing engine → Hive local database → machine-learning models (expense categorization and failure prediction) → budget monitoring engine → Flutter dashboard.

08

WORKFLOW

  • ▸ The app reads incoming bank SMS (with permission) and filters UPI messages.
  • ▸ The NLP engine cleans the text and extracts amount, merchant, bank, time and status.
  • ▸ Structured transactions are stored locally in Hive.
  • ▸ A machine-learning model categorizes each expense automatically.
  • ▸ The budget engine recalculates spending and warns at 80% and 100% of a limit.
  • ▸ A failure-prediction model estimates the probability of a UPI transaction failing.
  • ▸ Analytics, charts and the AI chat interface present the results.
09

SCREENSHOTS

Login
Dashboard
Dashboard detail
Transactions
Transactions detail
Analytics
Analytics detail
AI Center
AI financial chat
Profile
Settings
Running on a device — dashboard
Running on a device — SMS import
Running on a device — add transaction
10

MY CONTRIBUTION

Frontend developer. Built the app screens in Flutter, then connected the backend and adjusted it to work with the frontend.

11

CHALLENGES

  • ▸ Reading different bank SMS formats correctly.
  • ▸ Sorting expenses into the right categories automatically.
  • ▸ Connecting the backend to the frontend so both work together smoothly.
12

FUTURE IMPROVEMENTS

  • ▸ Version 2 for public release.
  • ▸ Bank API integration for real-time balance sync.
  • ▸ Credit score analysis.
  • ▸ Smart saving recommendations.
  • ▸ Investment suggestions.
  • ▸ Encrypted cloud backup.
  • ▸ Multi-language SMS parsing.