Diego est cofondateur et CTO d'anyformat. Il a étudié à l'Universidad Politécnica de Madrid et a travaillé dans la division GenAI de Clarity AI avant de cofonder anyformat. Il dirige l'équipe d'ingénierie qui construit l'architecture parse-extract d'anyformat, dont l'extracteur agentique, le runtime de workflows (Studio), l'opérateur Smart Lookup et la couche de visual grounding. Il écrit sur l'infrastructure documentaire, les protocoles comme MCP et les décisions d'ingénierie derrière l'extraction de niveau production.
Articles
PDF to Excel automatically: how to get PDF tables into a spreadsheet
· How to convert a PDF to Excel automatically without scrambled columns: what the built-in tools do, why multi-page tables and scans break, and how to check the numbers.How to extract data from bank statements automatically
· Extract transactions and balances from bank statement PDFs and scans automatically, with a balance check that catches wrong rows. Sample schema and JSON.How to extract data from long PDFs with an LLM
· How to extract data from long PDFs with an LLM without losing rows: why one prompt fails, multi-page tables, and how to check nothing is missing.Invoice data extraction API: how to extract invoices to JSON
· How to extract invoice data to JSON with an API: header fields, line items, confidence and evidence per value. Working code in curl and Python.OCR for poor-quality scanned PDFs
· A bad scan ruins OCR for a few predictable reasons. A checklist by problem, what we measured, and when to stop fixing the image and use an agentic IDP.The 7 Best Unstructured.io Alternatives in 2026
· The best Unstructured.io alternatives in 2026, open source and managed, compared on accuracy, field-level confidence, self-hosting, connectors and price per page.Best Python OCR libraries: Tesseract, EasyOCR, PaddleOCR
· We ran Tesseract, EasyOCR and PaddleOCR on the same synthetic images: clean, low-res, rotated, noisy, Spanish and a table. Scores, setup and where each breaks.Extract tables from a PDF: pdfplumber vs Camelot vs PyMuPDF
· We ran pdfplumber, Camelot and PyMuPDF on the same PDFs: ruled, borderless, merged cells, two pages and a scan. What each returns and where each breaks.Is Tesseract accurate enough for production?
· Tesseract publishes no accuracy figure. What its docs say, what independent tests found, and a 50-document test with a go or no-go threshold.OCR API vs open source: how to decide
· A break-even formula, today's published OCR API prices and a checklist to decide between a managed OCR API and self-hosted open source.The 7 Best Document Parsing APIs in 2026
· Seven document parsing APIs compared on published price, deployment, compliance and what comes back beyond markdown, with every price dated 30 September 2026.Document parsing for AI agents with MCP
· How to let an AI agent read PDFs through MCP: what a tool call can carry, how providers expose agent access, and a walkthrough to connect a coding agent and parse a document.Extract structured data from a PDF with an LLM
· A working Python walkthrough: read a PDF, define a schema, get typed JSON from an LLM and check it. Where the approach breaks and what to add for production.GDPR-compliant document AI: EU residency explained
· Residency, processing location and access are different questions. What EU data residency means for document AI, what eight vendors publish, and ten questions to ask any provider.Intelligent document processing vs OCR
· OCR turns an image into text. Intelligent document processing turns a document into checked, typed fields. The difference, with one invoice shown both ways.PDF to markdown for LLMs: Marker vs Docling vs MarkItDown vs anyformat
· We ran Marker, Docling and MarkItDown on the same PDFs: tables, two columns and a scan. What each one gets right, where each breaks, and how to test them on your own files.Document Packets: Many Files, One Document, Without the Glue Code
· A contract arrives with two annexes. An invoice arrives with the delivery note that justifies it. The document is the bundle, not any one file. A document packet is how anyformat runs a workflow on that bundle: one upload, one run, one set of extracted fields, with the context of every file in view.Flash Mode: Born-Digital PDFs, Without the Model Call
· Most of the PDFs a back office processes were never scanned. The text is already in the file. Flash is the Parse mode that reads it straight from the text layer, with no model in the loop, at 7 credits a page instead of 25, and returns the same blocks, bounding boxes and confidence every other operator expects.The 6 Best Rossum Alternatives in 2026
· Looking for a Rossum alternative? We compare 6 tools on pricing transparency, custom fields without a labeling project, on-premise deployment and EU sovereignty.The 9 Best Azure Document Intelligence Alternatives in 2026
· The best Azure Document Intelligence alternatives in 2026, compared on custom fields without training, self-hosting, EU data control, evaluation tooling and price.The 6 Best Nanonets Alternatives in 2026
· Looking for a Nanonets alternative? We compare 6 tools on zero-shot extraction, on-premise deployment, EU sovereignty, evaluation tooling and pricing.Knowledge Base: The Questions Your Schema Did Not Ask
· Structured output answers the questions you knew to ask when you designed the schema. Every anyformat workflow now ships with a knowledge base that answers the ones you did not, and every answer cites the page and the region it came from.LlamaParse vs Unstructured vs Reducto: Which One Fits Your RAG Pipeline?
· A field-by-field comparison of the three AI-native tools teams reach for first when building a RAG or document-ingestion pipeline: LlamaParse, Unstructured and Reducto, plus where anyformat fits if the pipeline needs to end in a business system, not a vector store.The 7 Best LlamaParse Alternatives in 2026
· Looking for a LlamaParse alternative? We compare 7 tools on review workflows, confidence you can act on, air-gapped deployment and pricing.Edit: Filled Forms, Without the Field Mapping
· Document intelligence has always run in one direction: documents in, data out. Edit is the operator that writes back. Hand a workflow a blank form and get the completed PDF, with every field detected for you and nothing mapped by hand.AWS Textract vs Google Document AI: Which One to Choose in 2026
· A neutral, numbers-first comparison of AWS Textract and Google Document AI in 2026: measured OCR quality, pricing, page limits, custom extraction and EU residency.The 8 Best AWS Textract Alternatives in 2026
· Looking for an AWS Textract alternative? We compare 8 tools on custom fields without training, on-premise and air-gapped deployment, EU sovereignty and pricing.The 8 Best Google Document AI Alternatives in 2026
· The top Google Document AI alternatives in 2026, compared on custom extraction, EU data sovereignty, on-premise deployment, evaluation tooling and pricing.The Best OCR APIs for Spanish-Language Documents (2026)
· We compare seven OCR and data extraction APIs for Spanish-language documents: anyformat, Google, Azure, AWS Textract, ABBYY, Tesseract and Nanonets.Evals: Change Your Document Workflows Without Fear
· Every anyformat workflow now has a Health tab: build a dataset with known-correct answers, score any workflow version against it, and see exactly what got better and what broke, before production finds out.Meet Annie, Your AI Doc Assistant
· Building workflows used to mean configuring every field by hand. Meet Annie: describe what you need in plain language, and she sets up the workflow and tunes it against your data. Setup drops from about 4 hours to around 5 minutes.Your AI Stack Can Disappear Overnight. Now What?
· The US government just forced Anthropic to pull its two most powerful models offline. Everyone's saying 'diversify your providers.' That's necessary but not sufficient — and here's why.Smart Lookup: Reference Data, Without the Brute Force
· Document intelligence is not just document parsing. Smart Lookup is the operator that turns reference data workflows from context-window gambling into structured, traceable queries.anyformat Studio: Complex Document Workflows, No Complexity
· Studio is the canvas where document intelligence pipelines become visible, composable, and accountable. Today it's live, and this is what it changes.ISO 27001:2022, Certified. The Trust Was Engineered Before the Audit.
· anyformat is now ISO 27001:2022 certified. The controls, the ISMS, and the architecture were engineered first. The certificate, audited by Prescient Security, confirms what was already in place.Delve Got Caught Faking Compliance. We Chose the Slow Way on Purpose.
· The Delve scandal is exposing what happens when compliance becomes a product to ship fast rather than a promise to keep. At anyformat, we took the opposite path, and it's taking us months. On purpose.OpenClaw Is Exciting. Your Documents Deserve Better Than Excitement.
· The viral AI agent reveals what happens when autonomy outpaces architecture, and why document intelligence demands a fundamentally different approach.AI Agents Don't Kill Document Processing. They Make It Inevitable.
· There's a narrative that agents and LLMs will make documents obsolete. I think that's fundamentally wrong. Here's why document intelligence becomes the substrate layer for every autonomous system.Model Context Protocol (MCP) and the AI-Native Era of Unstructured Data
· MCP is not “just another integration standard.” It fundamentally changes how AI interacts with unstructured data, turning documents into agentic conversations.
Commencez par vos documents les plus difficiles.
anyformat fait le gros du travail sur les documents qui font échouer d’autres outils. Parsez, extrayez et validez-les en données propres et fiables, et passez en production en quelques minutes.
Sans carte bancaire · 50 000 crédits gratuits pour commencer
