Technology
A modular, secure architecture hosted in Europe
AIR-Core combines a web application and its API with specialized modules for modelling, explainability, storage and AI. Everything runs in the European Union.
Architecture
How AIR-Core is built
Each module does one thing and does it well. That way we can scale, update and secure each piece separately.
Browser
Web interface to create sources, models and dashboards.
AI apps
Claude, ChatGPT and others, connected through MCP with OAuth.
Your systems
Integrations with the REST API v1, the Python SDK and webhooks.
AIR-Core · AWS Frankfurt (EU)
Web and API
Interface, REST API v1, two-factor authentication, organizations, roles and permissions, dashboards and alerts.
MCP server
Exposes sources, models and dashboards to AI apps with read or write permissions. It logs every call and offers no delete tools.
Modelling engine
AutoML in Python with scikit-learn and XGBoost: data preparation, algorithm selection and parameter tuning.
Explainability
Local and global explanations with SHAP values, and model cards. No generative AI.
Storage
Stores the sources and serves the queries of visualizations and models.
Generative AI
Claude on Amazon Bedrock with an EU inference profile, or the provider of your own key (BYOK).
Secrets
Database credentials and API keys encrypted with a master key.
Databases
MySQL, PostgreSQL, SQL Server, Oracle, MongoDB, InfluxDB.
Web and open data
URLs, APIs and CKAN portals.
Devices
Real-time data through MQTT.
Machine learning
Proven algorithms, chosen for you automatically
AIR-Core's modelling engine is built on open, well-tested Python libraries such as scikit-learn and XGBoost.
Classification and regression
Decision trees, ensembles (random forests, gradient boosting, XGBoost), linear and logistic regression, support vector machines, nearest neighbours, naive Bayes and neural networks.
Clustering
Grouping of similar cases without prior labels.
Anomaly detection
Methods such as Isolation Forest to flag unusual cases.
Association rules
Discovery of patterns like “if A, then B”.
Text (NLP)
Text classification and topic modelling on free-text columns.
Automatic selection
AIR-Core compares candidates, tunes their hyperparameters and picks the model that generalizes best.
Data preparation
From scattered data to useful information
Many organizations pile up data in formats that don't fit together. AIR-Core gathers them and prepares them for analysis.
- 1
Collection
From complex databases to text or CSV files.
- 2
Cleaning
Useful data is identified and irrelevant data discarded.
- 3
Combination
Sources are joined to cross-reference information.
- 4
Preparation
Formats and types are adapted for the algorithms.
- 5
Visualization
Simple, intuitive views for better decisions.
Visualization
The right format for each piece of data
Choosing how to show a piece of data matters as much as the data itself. AIR-Core offers everything from classic charts to maps, treemaps, chord diagrams and real-time series.
Security and data residency
Your data, in Europe and under your control
Technical and organizational measures appropriate to the risk, described in annex II of the data processing agreement.
Data in the EU
Hosting in AWS Frankfurt (eu-central-1). AIR-Core's AI runs in AWS regions in the EU.
Encryption
Traffic encrypted with TLS. Credentials and API keys encrypted at rest with a master key.
Access
Organizations, roles and permissions per workspace. Two-step verification.
Traceability
Log of account activity and of every use of AI, visible to administrators.
Backups
Regular backups of the platform to recover from incidents.
GDPR
We act as a data processor under a data processing agreement. We don't use your data to train models.
Start today with your own data
Create your free account and connect your first source in minutes. When you need more, move to a paid plan.