04 / Project
Lumen
An intelligent personal finance analysis application.
2025 · Software Engineer · Application
Overview
Lumen is a private analysis layer for personal finance — not a marketplace, not a feed. It reads structured transaction history and returns clear pictures of spending, saving, and drift over time, written in language a person can sit with.
The Problem
Most finance products either overwhelm with charts or reduce a household’s money to a score. People who wanted to understand patterns — not be gamified — had few tools that treated their data with the same care as a well-kept ledger.
The Approach
The product is organised as a sequence of questions: where did it go, what changed, what is repeating. Visualisation is secondary to prose summaries grounded in the numbers. Categories are editable. Nothing is sold inside the interface. Privacy is a constraint on architecture, not a footnote.
Implementation
Ingestion and classification run in a Python service with explicit, reviewable rules before any model assistance. The Next.js application presents monthly and yearly views from PostgreSQL aggregates. Sensitive fields are encrypted at rest. There is no advertising surface and no social graph.
Outcome
Users can read a year of activity as a short brief rather than a puzzle of graphs. The system surfaces drift early — subscriptions, seasonal spikes, quiet leaks — without turning money into a performance.
Technology
Next.js · TypeScript · Python · PostgreSQL
