Every business on the planet wastes hours copying numbers from documents into software. We decided to build the solution that nobody else was building: affordable, private, and fully on-premise.
This is the story of DataUnchain — from a single developer's frustration to a product that is redefining how small and medium enterprises handle document data.
DataUnchain exists for one reason: to give small and medium businesses the same document automation power that Fortune 500 companies enjoy — without the Fortune 500 budget, without sending sensitive data to the cloud, and without needing a team of engineers to set it up.
Our mission is straightforward. We believe that an invoice arriving via email should automatically become a record inside your ERP, your accounting software, or your management system. No human should have to open a PDF, read a number, switch to another window, and type that same number again. That workflow is a relic of the 1990s, and yet in 2026 it is still the daily reality for millions of businesses around the world.
We are building an AI-powered document ingestion pipeline that runs entirely on your hardware. It reads invoices, delivery notes, orders, receipts, and other structured documents. It extracts every relevant field — supplier name, VAT number, line items, totals, due dates — with an accuracy that matches or exceeds human operators. And then it pushes that data directly into your existing software through a growing library of connectors.
The result? A process that used to take eight hours per day now takes thirty seconds. A process that used to generate a 3-5% error rate now runs at 95.5% field-level accuracy on real-world documents. A process that used to require one or two full-time employees can now run unattended, overnight, while you sleep.
And it all happens inside your office walls. Your data never leaves your network. Not a single byte is sent to a third-party cloud. That is not a marketing claim — it is an architectural decision baked into the core of the product. DataUnchain was designed from day one to be 100% on-premise, and that is exactly what it will remain.
Developer. Builder. Obsessed with automation.
DataUnchain was not born inside a corporate innovation lab. It was not spun off from a university research project. It was not assembled by a team of twenty engineers with a Series A in their pockets. It was built from scratch by one person, sitting at a desk, writing code, testing models, debugging pipelines, and iterating relentlessly until the product worked.
Antonio Trento is a developer based in Italy with a deep background in web technologies, data engineering, and AI integration. Over the years he has worked across the full stack — from frontend interfaces to backend APIs, from database design to server infrastructure, from machine learning models to production deployment. His professional portfolio at antoniotrento.net reflects the breadth of that experience.
But DataUnchain is his most ambitious project yet. The idea originated from a simple observation: every small business Antonio worked with — accountants, logistics firms, import-export companies, healthcare clinics — was drowning in the same problem. Stacks of invoices and documents arriving every day. Employees manually copying data from those documents into their management software. Errors creeping in. Time evaporating. And when he looked for solutions, the market offered only two options: expensive cloud platforms that required sending sensitive data abroad, or enterprise products with six-figure price tags and twelve-month implementation timelines.
Neither option worked for a 15-person accounting firm in Milan or a family-owned logistics company in Puglia. So Antonio decided to build the third option himself.
The development journey has been intense. Antonio designed and built the entire architecture: a modular Python backend, a proprietary Vision Language Model fine-tuned specifically for structured document extraction, a multi-channel intake system (email, FTP, API, folder watch, web upload), a connector framework for ERP and accounting software, and a quality validation layer that catches errors before they reach the target system.
The benchmark results speak for themselves. On a test set of 219 real-world Italian business documents — invoices with irregular layouts, handwritten annotations, multi-page structures, and poor scan quality — DataUnchain achieved 95.5% field-level extraction accuracy. That number was not achieved on clean, synthetic test data. It was achieved on the messy, complicated documents that real businesses actually deal with every day.
The document processing market is broken in three fundamental ways. DataUnchain was built to fix all three at once.
The average European SME spends over €10,000 per year on manual data entry — when you factor in salaries, error correction, tax penalties from mistyped numbers, late payment fees from processing delays, and the massive opportunity cost of having skilled employees doing mindless copy-paste work.
Most businesses don't even realize the true cost because it's distributed across dozens of small inefficiencies. They see the salary line item. They don't see the €3,000 in correction time, the €2,000 in penalties, or the €10,000+ in opportunity cost. DataUnchain makes all of those costs disappear.
Every time you upload an invoice to a cloud OCR service — whether it's AWS Textract, Google Document AI, or Microsoft Azure Form Recognizer — you are sending your clients' names, VAT numbers, financial details, and business relationships to a data center in another country.
For businesses handling health records, legal documents, or financial data, this is not just a privacy concern — it's a regulatory minefield. GDPR, NIS2, sector-specific compliance frameworks — all of them require you to know exactly where your data goes. With cloud services, that answer is always "somewhere else." With DataUnchain, the answer is always "right here, on your server."
The products that actually work well — ABBYY, Kofax, Hyperscience — cost €50,000 or more per year. They require dedicated implementation teams, months of customization, and ongoing consulting fees. They were designed for banks and insurance companies processing millions of documents, not for accounting firms processing thousands.
The result is a giant gap in the market. Below the enterprise tier, there is essentially nothing that combines high accuracy, real privacy, and affordable pricing. DataUnchain was built specifically to fill that gap — to deliver enterprise-grade extraction at a price point that a 10-person company can afford.
DataUnchain is not a mockup or a slide deck. It is a working product with measurable results. Here is exactly where we stand right now.
Five input channels means documents can arrive via email (IMAP/POP3 monitoring), FTP/SFTP drop folders, REST API calls, local watched directories, or direct web upload. You don't need to change how suppliers send you invoices — DataUnchain adapts to your existing workflow.
Eighteen connectors means the extracted data can flow directly into the software you already use: Fatture in Cloud, TeamSystem, Danea Easyfatt, Zucchetti, SAP Business One, Odoo, custom databases via ODBC, and more. Each connector is a modular adapter that maps extracted fields to the target system's schema.
Seven document types currently supported: invoices, credit notes, delivery notes (DDT), purchase orders, receipts, pro-forma invoices, and self-invoices. Each type has its own extraction template optimized for the specific fields that matter.
Thirty seconds is the average end-to-end processing time per document — from intake to validated, structured JSON output ready for connector dispatch. That includes VLM inference, field validation, confidence scoring, and error detection. Compare that to the 3-5 minutes a human takes per invoice, and you begin to see the scale of the efficiency gain.
DataUnchain is at the beginning of an ambitious growth trajectory. Here is the roadmap — transparent, measurable, and grounded in reality.
DataUnchain has been submitted to Invitalia's Smart&Start Italia programme — the Italian government's primary funding instrument for innovative startups. The grant application covers €680,000 in subsidized financing for product development, market launch, infrastructure, and team expansion.
This is not venture capital. It is a government-backed programme designed to support Italian innovation. The funding structure provides zero-interest financing with a 30% non-repayable grant component, making it an ideal launchpad for a deep-tech product like DataUnchain.
The initial go-to-market targets 25 paying customers within the first 18 months of commercial launch. The primary segments are accounting firms (studi commercialisti), logistics companies handling high volumes of delivery notes and invoices, and healthcare organizations dealing with structured patient documentation.
At an average annual recurring revenue of €2,400 per customer, this translates to €60,000 ARR within the first 18 months — a solid foundation for a bootstrapped, capital-efficient business model.
Italy is the starting point, not the destination. The Italian market is ideal for a first launch because of its unique combination of high document volumes (electronic invoicing has been mandatory since 2019), strong privacy awareness, and a fragmented SME landscape that is underserved by existing solutions.
Once DataUnchain is established as the reference solution for Italian SME document automation, the architecture is designed to scale across Europe. The extraction model is language-agnostic — it works on document structure, not language. Adding support for German, French, Spanish, and Portuguese invoices requires only field-mapping configuration, not model retraining. The European market for SME document automation is valued at over €4 billion, and no dominant on-premise player exists.
We know what you're thinking. A solo-founder startup with no funding round and no enterprise client list — why should you bet on us? Here are five concrete reasons.
DataUnchain is not vaporware. It is not a landing page collecting emails while the founder figures out what to build. Version 2.1 is a fully functional pipeline: document intake from five channels, VLM-powered extraction using our proprietary Vision Language Model, field validation with confidence scoring, and output to eighteen ERP/accounting connectors. You can install it today and process your first invoice in under an hour.
We do not hide behind vague accuracy claims. Our benchmark methodology, dataset composition, and results are published and reproducible. 219 real Italian business documents. 95.5% field-level accuracy. Tested against Google Document AI, AWS Textract, and Azure Form Recognizer on the same dataset. These are not cherry-picked numbers — they are the unvarnished results of a rigorous, reproducible evaluation.
The core extraction engine is open source. You can audit the code, verify our privacy claims, and confirm that no data is sent to external servers. Open source is not a marketing tactic — it is a trust mechanism. When we say "100% on-premise," you don't have to take our word for it. You can read the source code and verify it yourself. The enterprise features, connectors, and management dashboard are available under a commercial license.
We are so confident in the product that we offer early adopters a full six-month trial period with no payment required. Not a limited demo. Not a feature-gated free tier. The complete product, with all connectors and all document types, running on your hardware for six months. If it doesn't pay for itself within that window, you walk away with zero cost and zero obligation.
When you sign up as an early adopter, you don't get a chatbot or a tier-1 support agent reading from a script. You get direct access to Antonio Trento — the person who designed, built, and maintains every component of the system. If something breaks, the person who wrote the code is the person who fixes it. That level of support is impossible at scale, which is exactly why it's so valuable right now.
DataUnchain is at the stage where the right people joining at the right time can change the entire trajectory of the company. Here is who we're actively looking for.
Accounting firms, logistics companies, healthcare organizations, or any SME processing 500+ documents per month. You get six months free, direct founder support, and the ability to shape the product roadmap based on your real-world needs. Your feedback is more valuable than any market research report.
A senior developer or ML engineer who wants to co-own a deep-tech product from the ground floor. Ideal profile: strong Python skills, experience with ML/NLP pipelines, and the hunger to build something that matters. You would work directly with Antonio on model optimization, connector development, and scaling the architecture for enterprise deployments.
IT consultants, system integrators, or software resellers who already work with SMEs and want to add a high-margin, sticky product to their portfolio. DataUnchain's on-premise model means recurring installation, configuration, and support revenue for partners — not just a referral fee. We are building a partner programme with generous margins and full technical enablement.
We are open to conversations with angel investors who understand deep-tech, B2B SaaS, and the European SME market. We are not looking for growth-at-all-costs capital. We are looking for patient, strategic investors who value capital efficiency, product quality, and sustainable unit economics. The Smart&Start grant provides the runway — angel investment would accelerate go-to-market and team building.
Every company claims to have values. We'd rather show ours through architectural decisions and business practices than through posters on a wall.
We publish our benchmark methodology. We open-source our core engine. We share our accuracy numbers — including the cases where we fall short. We tell you exactly what the product can and cannot do. We believe that trust is built through radical honesty, not through marketing spin.
When our model fails on a document, we don't hide it. We log it, analyse it, and use it to improve the next version. Transparency is not comfortable, but it is the only way to build a product that people genuinely trust with their business data.
Privacy is not a feature we added after launch. It is the foundational architectural decision that shaped every other design choice in the product. DataUnchain runs entirely on your hardware. Our proprietary VLM runs locally. No API calls to external servers. No telemetry. No usage data collection. Zero bytes leave your network.
This is not a checkbox for GDPR compliance — it is a genuine belief that businesses should own their data completely and unconditionally. Your invoices contain your clients' names, your pricing, your margins, your relationships. That information belongs to you, not to a cloud provider's training dataset.
We would rather ship one feature that works flawlessly than ten features that work "mostly." Every connector is tested against real production data. Every extraction template is validated against hundreds of document variants. Every release goes through a regression suite that checks accuracy across the entire benchmark set.
In document processing, "good enough" is not good enough. A 2% error rate on 10,000 invoices per month means 200 errors — 200 potential tax penalties, 200 potential payment disputes, 200 reasons for a customer to lose trust. We measure quality in errors per thousand documents, and our goal is to drive that number as close to zero as the laws of physics allow.
The entire reason DataUnchain exists is that enterprise-grade document AI was inaccessible to small and medium businesses. Our pricing, our deployment model, our documentation, and our support are all designed around one principle: if you need this product, you should be able to afford it and use it, regardless of whether you have an IT department.
We believe that the democratization of AI is not about giving everyone access to a chatbot. It is about giving every business access to the automation tools that were previously reserved for the largest and wealthiest organizations. DataUnchain is our contribution to that mission.
Whether you're an SME tired of manual data entry, a developer who wants to co-build the future of document AI, or an investor looking for a capital-efficient deep-tech opportunity — we want to hear from you.