OUR TECHNOLOGY
Configurable and extensible pipeline system.
Metamaze is an intelligent solution that can handle diverse types of documents with a high degree of accuracy and minimal manual oversight.
The brains in the background.
Winning organizations process and interpret documents faster than their competition because they back away from outdated and slow manual document processes causing delays and frustrations. Future-fit organizations choose an adaptive Intelligent Document Processing solution, like Metamaze.
OUR TECHNOLOGY
Faster onboarding by focusing on the annotation process first.
Optimal Document Selection Strategy
Auto-correct for annotation mistakes
Fully customizable pipeline
- Easily add document types, entities, enrichmentss or validation rules.
- Build models from scratch or fork from our continuously expanding model repository
- Validate and enrich data with business rules and by connecting to your own IT systems
- Customize pipeline behavior with custom code snippets using Enrichments.
Clutter-free, simple user interface for daily operations
AI operators see only what they have to see: an overview of the work they can pick up, along with everything that needs to happen for that document. This is true for production, training, QA tasks, and anywhere else.
Queues are real-time and can be filtered on language, document type, custom metadata, document name, ID, upload date, or assigned user.
If an operator gets stuck during annotation or validation, documents can be parked with a comment for further assistance or review by managers.
Clear production and model analytics
Metamaze has detailed production analytics with so your team can continuously measure and further improve quality and automation rate.
Model analytics give deep insight into metrics calculated during the training of the model, like expected STP rate, f1 score, precision, recall for all regular entities and composite entities. Precision/recall curves can be used to optimally define the confidence threshold. Confusion matrices can be used to see if it would make sense to merge document types and where the model still makes mistakes.
Feature details
Data and model management
Track, compare, monitor and optimize model development, from training all the way to production.
Fast & easy labeling
Manage data labeling guidelines to facilitate the data entry users with step-by-step instructions.
Human intervention configurable
Manage quality threshold and determine human-in-the-loop processes all by yourself.
Unique blend of AI and humans
We use feedback from human-in-the-loop processes as input data for the deep learning models.
User roles
Advanced role and user management capabilities for proper feature allocation and protection from accidental misuse or abuse.
Queueing
Real-time overview of what is happening and possibilities to collaborate with other team members.
Audit trail
Any activity on the platform, human or AI, will be logged within Metamaze's detailed audit trail.
Monitoring
Real-time dashboards that monitor performance (cost, savings, ...) and various parameters (STP rate, data quality, ...).
