Beyond the cloud monolith: Building transparent and distributed AI with open source

"Beyond the cloud monolith: Building transparent and distributed AI with open source," presented by Raphaël Semeteys (Worldline), explores how open source enables a shift from centralised AI systems to local, distributed, and decentralised architectures. This session is part of the Open Community for AI at Open Community Experience 2026 in Brussels, Belgium.

This talk challenges the dominance of hyper-centralised AI by examining alternative architectures built on open source principles. It outlines how current AI ecosystems are shaped by a concentration of power, creating dependencies, sustainability concerns, and geopolitical implications.

A core part of the session defines what “open” means in AI, distinguishing between open weights, access to training data, and reproducibility of models. It highlights that most of today’s landscape sits in an “open weight” model, where models can be reused and adapted, but full transparency is limited by restricted access to datasets and training pipelines.

The talk then breaks down the local inference stack, covering hardware layers, acceleration frameworks, inference engines, and model optimisation techniques such as pruning, quantisation, and distillation. These enable smaller, specialised models to run on local infrastructure, reducing reliance on centralised cloud systems and supporting more efficient, domain-specific AI deployment.

Building on this, the session explores distributed AI architectures, including orchestration frameworks, agent-based systems, and interoperability protocols that allow multiple models and services to work together. It examines how inference and workloads can be distributed across systems to overcome hardware constraints and improve scalability.

Finally, the talk introduces decentralised AI, where control, verification, and value distribution are handled through networked systems rather than central authorities. This includes decentralised compute marketplaces, data governance models, distributed training, and agent-based economies, highlighting both the technical possibilities and the complexity of implementing such systems.

Key topics covered
- centralised vs distributed ai architectures
- open source ai and open weight models
- ai model transparency and reproducibility
- local inference stack and edge ai deployment
- model optimisation techniques (pruning, quantisation, distillation)
- token-based incentive systems in ai networks
- decentralised validation and trust mechanisms
- future of ai beyond cloud-centric models

Why this matters
AI is increasingly centralised in a small number of platforms, creating technical, economic, and sustainability risks. Open source and distributed approaches offer alternative models that improve resilience, reduce dependency, and enable more transparent and adaptable AI systems.

About OCX26
Open Community Experience 2026 is the Eclipse Foundation’s flagship event, held in Brussels, Belgium. It brings together developers, architects, and industry leaders to explore open source technologies across domains including AI, automotive, tooling, and cloud systems, with a focus on practical implementation. Learn more at https://www.ocxconf.org/

Chapters
00:00 introduction and centralised ai critique
02:16 risks of ai concentration and dependency
03:36 open source as an alternative model
04:01 defining openness in ai models
05:50 model weights, data, and reproducibility
08:22 current landscape of open weight models
11:21 benefits of openness and local ai
12:44 local inference architecture overview
13:50 hardware and acceleration layers
15:39 inference frameworks and execution layers
17:18 model optimisation techniques
19:09 model serving and inference systems
21:31 user interfaces and developer tools
23:07 orchestration and ai workflows
25:27 agent-based systems and multi-agent architectures
27:51 distributed ai architectures
28:41 distributed inference and scaling
31:36 decentralisation vs distribution
35:36 decentralised ai ecosystems overview
36:05 decentralised compute marketplaces
37:13 decentralised data and governance
38:59 decentralised intelligence and training
41:34 decentralised validation mechanisms
42:38 autonomous agents and ai economies
44:49 sustainability and future outlook

Видео Beyond the cloud monolith: Building transparent and distributed AI with open source канала Open Community Experience