FinPal: An App That Answers Real Questions About Your UPI Spending

FinPal is a personal finance app built around India's UPI payment ecosystem, automatically categorizing transactions from GPay, PhonePe, and bank CSV/PDF exports, then letting users ask plain-English questions about their own spending. Citing NPCI figures showing billions of monthly transactions and hundreds of millions of active users, the project argues that when financial life becomes hundreds of small scattered UPI pings rather than a few large bank entries, a simple transaction list stops being enough — you need something you can actually query.

The technical core is a hybrid engine: a fast, cheap rules-based categorizer handles most transactions, escalating to Google Gemini only for ambiguous merchant strings the rules can't confidently resolve. The "Ask FinPal" chat layer takes this further by feeding Gemini a compact structured summary of the user's categorized spending — not raw transaction dumps — so answers to questions like "is that more than usual?" are grounded in real computed averages rather than generic financial advice.

Built on a React/TypeScript frontend and Node.js/Express backend, with Multer, csv-parser, pdf-parse, and xlsx normalizing inconsistent Indian bank export formats into one schema, FinPal's core engineering lesson is about selective AI use: keeping the fast path fast and reserving LLM calls for cases — ambiguous categorization and open-ended natural-language queries — where they genuinely add value, rather than routing every request through a model.