Physical AI for the Industries That Move the World

17+ years building automotive AI at Renault-Nissan-Mitsubishi Alliance and through Auralink. Deputy GM for NissanConnect — deployed across 39 countries and 4M+ users, including the world-first Google-built-in Android Automotive head unit (OpenR Link). Built Auralink: 400+ microservices and ~20 AI agents for EV charging orchestration, BESS coordination, and grid balancing (arXiv preprint 2603.08736). Vectis AI adds OBD-II and ESP32 vehicle edge inference. We work to ISO 26262, ISO 21448 SOTIF, UNECE R155 cybersecurity, UNECE R156 OTA, and AUTOSAR.

Manufacturing AI fails because the architecture is wrong, not the models. Auralink's IIoT Modbus bridge and ROS 2 multi-agent orchestration architecture demonstrate exactly the OT/IT integration pattern that factory-floor Physical AI requires. We help manufacturers build production-grade AI with proper edge deployment, ISA-95-aligned OT integration, and MLOps — the infrastructure that keeps models running after the engagement ends. We work to IEC 61508 SIL, IEC 62443 OT cybersecurity, and ISO 10218 robot safety standards.

Auralink delivers the strongest energy-sector proof in our portfolio: distributed EV charging network AI with 78% autonomous incident resolution, real-time BESS and solar coordination, OpenADR demand-response integration, and grid-balancing across distributed charging assets (arXiv preprint 2603.08736). ABB E-mobility A400 (400 kW ultra-fast charging) rounds out the infrastructure depth. We bring IEC 61850, IEC 62443, NIS2, and OpenADR discipline to grid-edge AI — not as a compliance add-on but as the architecture constraint from day one.

Modern logistics runs on Physical AI: autonomous mobile robots, computer-vision receiving lines, AI-directed yard and dock operations. The challenge is not individual automation islands — it is the distributed-agent architecture that coordinates them under real-time constraints, with on-prem inference and no cloud dependency. Auralink's multi-agent resource-arbitration architecture — proven in coordinating distributed EV charging assets under grid-balancing constraints — provides directly transferable primitives for warehouse AMR/AGV fleet orchestration. We work to ISO 3691-4 (driverless industrial trucks) and IEC 62443 OT security for logistics edge deployments.

Every AI system deployed by a public authority is high-risk under EU AI Act Annex III. That means conformity assessments, human oversight mechanisms, transparency obligations, and detailed audit logs before a single citizen interaction. We help civil agencies and sovereign-tech programmes deploy AI that is both effective and legally defensible — on controlled, EU-resident infrastructure, with the Mistral open-weight sovereign stack where data residency and air-gap requirements apply. The French Government AI Ambassador credential reflects direct engagement with AI governance and the EU AI Act in regulated environments. EU-AI-Navigator, our purpose-built EU AI Act compliance venture, provides the technical tooling layer.

Robotics AI fails at the boundary between simulation and the physical world. We design production-grade vision, manipulation, and coordination systems for cobot integrators, AMR vendors, and ROS 2 shops — deployed on-robot with no cloud dependency, built to ISO 10218 and ISO TS 15066 cobot safety requirements. Auralink's distributed-agent resource-arbitration architecture (arXiv 2603.08736) provides directly transferable primitives for multi-robot coordination.

Aerospace AI operates under the most demanding software assurance standards in any industry. We help MRO operators, avionics suppliers, and UAV makers deploy AI that is architecturally compatible with DO-178C, DO-254, and ARP4754A — from MRO copilots over technical manuals to automated NDT visual inspection and edge inference in avionics. Our focus is civil aviation and UAV; military programmes are out of scope.

Sovereign-tech and dual-use programmes require AI that runs entirely on controlled infrastructure — air-gapped, on-premise, with no data leaving the security perimeter. Our sovereign-first stack (Mistral open-weight → on-prem inference → edge sensor-fusion) is designed for exactly these constraints. We engage on the civil and dual-use technology layer: sovereign infrastructure design, on-prem model deployment, and EU AI Act dual-use compliance. We have no defence contracts, clearances, or government security accreditations — clients requiring those must engage cleared primes.

Space AI must operate autonomously under the harshest physical constraints in any domain: radiation-altered compute, minutes of latency to ground, and extreme SWaP-C budgets. We help satellite operators, Earth-observation newspace companies, and ground-station teams deploy AI that works on-board and on-ground — from EO image triage at the edge to constellation operations autonomy. Capability-only engagement: we bring transferable edge-AI and autonomy architecture expertise; we do not claim existing space-operator contracts.

Maritime AI must operate in RF-degraded, high-humidity, and mechanically harsh environments where cloud connectivity is intermittent at best. We help port operators, shipping lines, offshore asset managers, and autonomous-vessel makers deploy AI that works at the edge — from port AMR and crane automation to subsea robotics and remote predictive maintenance. Our distributed-agent and RF-spectrum primitives from Auralink provide transferable architectural foundations for port and vessel operations.