Building an AI platform on real-time network telemetry

Context

AXON Networks’ orchestration platform turns high-volume network telemetry into operational insight. Each device streams up to 50,000 datapoints per second, so the platform reads its network in real time rather than in batches, and every decision about data flow and model placement has to respect that latency budget. I lead the engineering and AI organisation for this in EMEA, around 40 people, and support integration with another 70 across Europe. The brief was to take a capable engineering setup and move it toward AI-driven operations without slowing the product down.

What I did

I defined the target cloud architecture and built a new AI/ML team from scratch, focused on LLM-driven agents that act on live telemetry rather than on stale snapshots. To make that possible we scaled the real-time data pipelines so the models read the high-frequency streams directly.

I set up a continuous-delivery culture with clear ownership so teams ship without waiting on a central bottleneck. Alongside the platform work I integrated a design and UX workflow and launched a third-party device team.

I also led the integration of the ACS system into the platform and supported platform integration after acquisitions, keeping one architecture coherent as new systems came in.

Outcome

The platform now runs real-time pipelines that feed operational intelligence and agent-based workflows on live telemetry. The AI/ML team is established and shipping. The architecture holds the latency the product needs, and the organisation has moved from a modern engineering setup toward AI-driven operations. The direction from here is set: agents take on more of the operational workflow as the data and models mature.