2026

AI Orchestrator, NVIDIA x Gcore webinar proto

Sovereign AI demonstration prototype, « Orange x Gcore AI Orchestrator », vibe-coded with Claude Code in under a day for the global NVIDIA x Gcore webinar (June 2026). Architecture vision: NVIDIA Dynamo for edge inference. Live.

proto-ai-webinar.vercel.app

AI Orchestrator prototype interface: catalogue of open models deployable on sovereign European infrastructure, with a Sovereignty Index — AI Orchestrator, the model catalogue with its Sovereignty Index

Overview

In June 2026, Orange Business took part in a global NVIDIA x Gcore webinar, working alongside Gcore as a partner. For that context, I built « Orange x Gcore AI Orchestrator », a sovereign AI demonstration prototype, vibe-coded with Claude Code in under a day. It is not a product: it is a demonstration piece that illustrates the architecture vision, with NVIDIA Dynamo for edge inference. The proto is live.

Sovereignty Index

Sovereignty step of the AI Orchestrator proto: a 94/100 score ring and four sovereignty gauges (data, technical, operational, legal), with an EU-isolated processing toggle. — The Sovereignty Index computed live inside the deployment wizard (demo data)

The proto scores each configuration on four sovereignty dimensions (data, technical, operational, legal) and recomputes the index live as options change. It is a demonstration piece, not a product: the scores are demo data.

Sovereign deployment

Infrastructure step of the deployment wizard: a map of Europe with cloud regions, a deployment footprint box and a summary panel listing model, regions, hardware and optimisation. — Wizard step 1: sites, hardware and configuration summary (demo data)

From the catalogue, an open model is deployed in three steps on European infrastructure: choice of sites, GPU hardware and allocation mode, with the summary and sovereignty index in a side column. Vibe-coded with Claude Code in under a day.

NVIDIA Dynamo

The architecture vision rests on NVIDIA Dynamo for edge inference: the Scaling step switches on the Dynamo optimisation and the routing through the Orange edge network before launching the deployment.

Edge inference

AI Grid screen of the proto: dark map of Europe with orange points of presence linked to Paris, a routed-tokens counter and the NVIDIA Dynamo inference engine badge. — AI Grid: token routing to the closest point of presence (demo data)

The AI Grid screen illustrates edge inference: the model runs in European cloud regions and tokens are served from edge points of presence, never leaving the EU. An architecture vision, presented in the context of the global NVIDIA x Gcore webinar of June 2026.

Live monitoring

A monitoring dashboard completes the demonstration: requests, latency, throughput, GPU utilisation, Dynamo KV-cache reuse, cost and carbon footprint. The proto is live at proto-ai-webinar.vercel.app.

Results

Prototype built with Claude Code in under a day
Presented in the context of the global NVIDIA x Gcore webinar (June 2026)
Edge-inference architecture vision built on NVIDIA Dynamo
Live at proto-ai-webinar.vercel.app

Technologies

Sovereign AINVIDIA DynamoClaude CodeReactInference

Next project