A European Research Commons

BEAM Campus

Open infrastructure for sovereign, distributed and evolutionary AI — built on the BEAM, running across a federation of small machines owned by no one and open to everyone.

Open source · European · built on the BEAM (Erlang/OTP)

What it is

AI that runs on many small machines — not one big cloud.

BEAM Campus is an open, European effort to build the infrastructure for distributed and evolutionary AI: a federated mesh of small, low-power nodes that anyone can join, with no central owner. The tools are open source and run on the BEAM — the Erlang virtual machine behind decades of resilient, highly concurrent, distributed systems.

The stack

Four open pieces, one substrate

Each layer stands on its own and is open source. Together they let populations of agents live, migrate and evolve across the mesh.

The network

Macula

A federated relay mesh over QUIC — DHT routing, membership and identity. Nodes connect across Europe with no single provider and automatic failover.

The data

ReckonDB

A BEAM-native event store with zero external dependencies — distributed event sourcing on Raft, with a full CQRS framework. Published on hex.pm.

The platform

Hecate

A sovereign application platform: a mesh-connected daemon plus a desktop app and an app store for extensions.

The evolution

Faber

The evolutionary layer — neuroevolution (topology-and-weight-evolving neural networks) on the BEAM, built to evolve populations across the mesh.

Why a mesh

The science and the substrate agree

Evolutionary methods evaluate many candidate solutions independently, generation after generation. There is no single enormous model to shard and no gradient to synchronise — just a population and its fitness. That work spreads naturally across many small, unreliable, geographically scattered machines: exactly what a federated mesh is good at, and exactly what a hyperscale cloud is not.

Get involved

Three ways to take part

Run a node

Contribute compute to the mesh and help the federation grow. Every node that joins makes the network stronger and more resilient.

Get in touch

Build on it

The core libraries are open source and on hex.pm. Integrate Macula or ReckonDB in your own Erlang/Elixir project, or dig into how it works.

Browse the code

Collaborate

Working on distributed or evolutionary AI, or curious where this is going? We'd like to hear from you.

Say hello