From signal to decision, in real time.

Arawise is a European data platform for organisations that work with sensitive data. Streams, tables and documents land in projects, you define an ontology of object types over them, and people and AI agents work on that model: SQL, notebooks, charts, alerts and automations, all under project-level access control. It is built on open standards such as Apache Iceberg, Trino and Kafka-compatible topics.

An ontology is a model of the things your organisation runs: object types such as Vessel, Berth or Aircraft, their fields, and the relations between them. Instead of asking analysts to rediscover the same joins in every query, the ontology captures them once. In Arawise you define the ontology yourself and the platform keeps it in sync with the tables underneath.

Each object type is mapped to tables in the lake: one table establishes which objects exist, and further sources and joins fill in fields. The sync runs whenever a mapped table receives a commit, so objects reflect the latest data without a separate pipeline to maintain.

Yes. Topics carry streaming data into a project, transforms shape it with streaming SQL, and alerts fire on records as they arrive. A stream can also be declared on an object type, so each object keeps its own record history, for example every position report for a vessel.

A project is the bounded unit for data and work. It has its own storage, its own SQL schema and its own access roles. Sources land in projects, flows and notebooks run in projects, and access is decided at the project boundary.

Agents scoped to a project can answer questions over its data and the ontology, render tables and charts, and read documents. A project assistant can draft notebooks, transforms and alerts on request. A person reviews and approves before anything is created.

Arawise is multi-tenant with a physically separated data plane for every tenant, hosted in Europe. Within a tenant, access is granted per project and per object type, and data markings follow records into every read, including reads by agents.