AI Invoices Processing - Cevital Agro-Industrie
Problem
Manual invoice processing in Transit. Slow. Error‑prone. Fragmented desktop tools. The company needed a modern web platform to automate payment requests with better efficiency.
Solution
Modern web platform to automate payment requests with AI‑assisted parsing of invoices and human‑in‑the‑loop validation.
Key Features
- Multimodal extraction with Gemma 3 mapped to domain entities (PaymentRequest, Invoice, Supplier, Contract)
- Batch capture per request with auto‑numbering and consistency via model observers (DR‑YYYY‑NNNNN, PR‑YYYY‑NNNN, supplier codes)
- Typed enums for currency and payment modes
- Fast UI with TypeScript + Tailwind, SPA UX via Inertia, server‑side rendering for snappy loads
- PDF exports aligned to enterprise templates (DomPDF + NumberToWords)
- Rich filters, search, and totals computed efficiently at the DB layer
- Seamless supplier/contract lookups and active‑contract guardrails
Impact
- Processing time: hours → minutes per batch
- Fewer errors and rework via structured capture + validation
- Clear traceability with state transitions (draft → submitted → approved → paid)
- Repeatable releases with Docker‑based deployment