The question behind the work
A bank statement contains plenty of data, but making it useful means getting the structure right first. This project brings document processing, validation and financial analysis into a single application.
DOCUMENT INTELLIGENCE
Make sense of the statement.
A multi-stage AI pipeline that turns bank statement PDFs into structured transactions, financial analytics and useful insights.
THE PROCESSINPUT → UNDERSTANDING
01Extract
Read transactions from statement PDFs.
02Validate
Check amounts, dates and transaction structure.
03Understand
Normalize merchants and categorize spending.
04Explore
Connect analytics, anomalies, goals and forecasts.
THE PROJECT
Personal Finance Agent
Source and context from my portfolio
How it comes together
The interface uses React and TypeScript, backed by Flask, PostgreSQL and SQLAlchemy. A policy-guided, multi-stage LLM pipeline supports extraction and analysis, with Ollama for local inference.
What to keep in mind
The useful insight comes after the data is structured and checked. Validation and traceability are central to this workflow.
From rows to a clearer picture.
Illustrative transactions, calculated in your browser. No statement upload or financial advice.
| Merchant | Category | Amount |
|---|
SAMPLE SPENDINGILLUSTRATIVE DATA