On This Page

Belgium-based IDP vendor acquired by Duco in February 2024, known for proprietary deep learning requiring 50% fewer training documents than competitors.

Overview

Metamaze was an Antwerp-based SaaS company spun off from AI consultancy Faktion in 2017, building adaptive intelligent document processing on proprietary deep learning that requires 50% fewer training documents than competitors to reach equivalent accuracy. Its Hydra model architecture addressed a specific gap in the IDP market: handling partially annotated training data, which most platforms struggle with. At acquisition, the platform processed over 400,000 documents weekly for a Belgian financial sector client and automated more than 40 document types.

Duco acquired Metamaze on February 6, 2024, for an undisclosed sum. The strategic logic was explicit: Duco's existing stack handled structured data reconciliation; Metamaze added unstructured document extraction. Together, Duco CEO Michael Chin argued, the combined platform eliminates the need for separate point solutions, letting financial institutions automate "manual data entry, manual data validation and reconciliation" in a single environment. Chin, who joined Duco the month before the acquisition closed, identified post-trade operations in capital markets as "under-automated to this day" and the primary go-to-market target.

Metamaze's transition from independent vendor to integrated platform component is complete. Whether its technology surfaces as a named product within Duco or disappears into the platform has not been disclosed. Buyers who evaluated Metamaze as a best-of-breed IDP point solution should note that its roadmap is now shaped by Duco's capital markets priorities, not the broader IDP market. Duco's claim of enabling a "90% reduction in data-related work" is aspirational; no benchmark or customer evidence has been published to support it.

How Metamaze processes documents

Metamaze's core differentiation was its machine learning approach to training data scarcity. The Hydra model architecture solved the partially annotated data problem, a common bottleneck where enterprises cannot label enough documents to train conventional models, by requiring 50% fewer annotated examples than competing platforms to reach production accuracy.

Beyond training efficiency, the platform operated as a self-learning system: extraction models improved continuously through live document processing without requiring manual retraining cycles or preset field configurations. CEO Niels Van Weereld described the design intent directly: "Most IDP platforms are limited by preset models, common document types, and fixed sets of fields. We've built Metamaze as an AI-powered self-learning intelligence platform that reads and interprets any document to extract information and data by itself."

The platform combined this adaptive machine learning core with active learning and human-in-the-loop validation, multilingual training, and API/webhook integration, delivered as a cloud-native, no-code SaaS. Business users could configure and operate document workflows without engineering involvement, which aligned with Duco founder Christian Nentwich's assessment: "in the cloud, with AI doing the heavy lifting, and focused on a powerful business end-user experience, based on a no-code framework." Vendors taking a comparable no-code approach to financial document automation include Unstract, which similarly targets production-grade extraction without requiring developer involvement.

Use cases

Financial services

Post-trade operations were Metamaze's strongest vertical at acquisition and remain the explicit focus under Duco. The platform demonstrated production scale at 400,000+ documents weekly for a Belgian financial institution, automating more than 40 document types including accounts processing and regulatory compliance workflows. Duco's acquisition rationale deepens rather than broadens this focus: the combined platform targets the specific, high-complexity workflows of capital markets operations teams, trade confirmations, reconciliation inputs, and related unstructured document flows, rather than general financial services automation. Alkymi addresses a comparable financial services extraction use case, specializing in transforming unstructured documents into standardized datasets for investment workflows.

Insurance

Banking and insurance were Metamaze's two named primary verticals at the time of acquisition. Insurance document automation, policy documents, claims inputs, correspondence, fit the platform's strength in handling variable document structures without preset templates. Paperbox, another Belgium-based IDP vendor, addresses a similar insurance mailroom automation use case and offers a point of comparison for buyers evaluating regional alternatives.

Air cargo operations

Prior to acquisition, Metamaze collaborated with CHAMP on Air Waybill data capture and comparison workflows, demonstrating applicability beyond financial services in logistics document processing.

Technical specifications

Feature Details
Architecture Cloud-native SaaS with proprietary deep learning
Training Efficiency 50% fewer annotated documents vs. competitors (Hydra model)
Processing Scale 400,000+ documents weekly (proven, Belgian financial client)
Document Types 40+ automated types
Integration API/webhook connectivity
Languages Multilingual training capabilities
Learning Model Active learning with human-in-the-loop validation
Deployment Cloud (SaaS)
No-Code Business user configuration without engineering

Resources

Company information

Metamaze operated as an independent vendor from its 2017 founding, as a Faktion spinoff, until its acquisition by Duco on February 6, 2024. Headquartered at Oudeleeuwenrui 39, Antwerp, Belgium, the company served banking and insurance clients primarily in Belgium and the broader European market.

Duco is a London-based SaaS provider of AI-powered data automation, backed by Nordic Capital since 2021. No integration timeline, headcount figures, ARR, customer count, or Metamaze brand continuity decision has been publicly disclosed. For current contact information and services, refer to Duco's official channels.