The latest articles on DEV Community by FLOW by Vestelon (@vestelonflow). https://dev.to/vestelonflow https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3991476%2F2cf94fec-5193-48fa-af66-aa548db3698a.png https://dev.to/vestelonflow en FLOW by Vestelon Thu, 18 Jun 2026 20:16:07 +0000 https://dev.to/vestelonflow/how-i-built-a-pdf-bank-statement-analyzer-in-8-languages-and-what-i-learned-3580 https://dev.to/vestelonflow/how-i-built-a-pdf-bank-statement-analyzer-in-8-languages-and-what-i-learned-3580 <p>I spent months building FLOW (<a href="https://vestelonflow.com" rel="noopener noreferrer">vestelonflow.com</a>) — a tool that analyzes bank statement PDFs and finds forgotten subscriptions, hidden fees, and recurring charges.</p> <p>Here's what I learned building it in 8 languages.</p> <h2> The Problem </h2> <p>Most personal finance apps require you to connect your bank account. For many people (especially in Europe), that's a dealbreaker. GDPR concerns, privacy fears, and simply not trusting third-party apps with banking credentials.</p> <p>My insight: <strong>the data people need is already in their PDF bank statements.</strong> Every bank generates them. Most people never look past the total.</p> <h2> The Tech Stack </h2> <p>The core flow:</p> <ol> <li>User uploads PDF bank statement</li> <li>PDF text extraction (pdfplumber + fallback OCR)</li> <li>Transaction parsing — this is the hard part</li> <li>LLM categorization pipeline</li> <li>Subscription detection (recurring charges with same merchant)</li> <li>Report generation</li> </ol> <p>The trickiest part was transaction parsing. Every bank formats their PDF differently. German banks look nothing like Slovak banks. We ended up building bank-specific parsers for the most common formats and a fallback generic parser.</p> <h2> The 8-Language Challenge </h2> <p>Supporting Slovak, Czech, German, French, Spanish, Polish, Arabic, and Chinese wasn't just about translating the UI. The financial terminology varies significantly:</p> <ul> <li>"Permanent order" in English = "Trvalý príkaz" in Slovak = "Dauerauftrag" in German</li> <li>Subscription detection keywords differ by region</li> <li>Date/amount formats are locale-specific</li> </ul> <p>We ended up with language-specific merchant dictionaries for common subscription services in each market.</p> <h2> What Actually Matters </h2> <p>The biggest lesson: <strong>people don't want a budgeting dashboard. They want a specific, actionable number.</strong></p> <p>"You're spending €137/month on forgotten subscriptions" converts. "Your spending breakdown by category" does not.</p> <p>The product is live at <a href="https://vestelonflow.com" rel="noopener noreferrer">vestelonflow.com</a> — first report is free, no card required, no bank connection needed.</p> <p>Happy to answer questions about the PDF parsing approach, the LLM pipeline, or the localization challenges.</p> fintech python showdev ai