See us at: Cisco GSX (Las Vegas · Aug 24–27)Cisco Connect Boston (Sept 23)Cisco Connect LATAM (Sept 28–30)Cisco Connect Atlanta (Sept 30)Cisco Connect Toronto (Oct 21) Book a meeting →
Cisco PartnerSenior engineers in North America & Europe · 24/7 escalation
SupportContact SalesENESPTFR
Home/Use Cases/Secure AI Factory

A stalled Secure AI Factory, validated in days — not months.

A major U.S. healthcare provider set out to prove that enterprise AI workloads could run securely on-premises — but integrating Cisco UCS, Red Hat OpenShift, Isovalent, Cisco AI Defense, NVIDIA GPUs and Splunk had stalled for months across siloed teams. Ngenium engineers stood the full stack up, working and observable, within days.

Days
to a validated Secure AI Factory — not the months the project had stalled
Delivered engagement
4
infrastructure layers unified: AI security, workload, compute & observability
Delivered engagement
1
integrated view across GPUs, workloads, security & model-validation events
Delivered engagement
PROVENANCE

The work described in this use case was delivered by Ngenium engineers, and reflects the hands-on Cisco Secure AI Factory expertise the team brings to every customer engagement. The customer is described by profile only and is not identified.

The challenge

Three priorities that had to be proven before broader adoption.

During discovery, Ngenium engineers and Cisco worked with the provider’s infrastructure, security, OpenShift and operations teams to translate broad AI infrastructure goals into concrete, measurable priorities.

01 — STALLED DEPLOYMENT

Months of delay across a multi-vendor stack.

The combination of Cisco UCS, Red Hat OpenShift, Isovalent Cilium, Cisco AI Defense and NVIDIA GPUs created an integration challenge that exceeded available bandwidth and cross-vendor expertise. Delays had accumulated across multiple workstreams, and the teams lacked a shared deployment path. The provider needed a partner who could cut through the complexity and stand up a working environment quickly.

02 — UNPROVEN WORKLOAD SECURITY

Visibility below the network layer, still on paper.

AI workloads, containerized applications and model inference environments required process-level security controls and behavioral telemetry that existing tooling could not provide. There were also open questions about whether Isovalent — a newer addition to the Cisco portfolio — was ready for enterprise scale. It had to be proven in practice, not asserted in a presentation.

03 — FRAGMENTED OBSERVABILITY

Six teams, six tooling silos, no shared view.

Compute, networking, storage, OpenShift, security and AI teams each worked from separate tooling. There was no unified view across GPU utilization, model-validation events, infrastructure health and workload behavior — and no reliable way to troubleshoot quickly or manage server lifecycle proactively across a large estate.

THE COMMON THREAD

Each challenge reflected the same underlying reality: Cisco’s Secure AI Factory represents a fundamentally new operating model — one that demands deep, coordinated expertise across infrastructure, container networking, AI security and enterprise observability, all working in concert. Very few internal teams anywhere hold that combination in-house. The evaluation demanded a partner who could deliver the full stack, prove it working, and connect it to the operational outcomes the provider’s teams needed to see.

The solution

One secure AI platform, from compute to workload.

Rather than four separate infrastructure projects, the provider’s teams saw a cohesive platform in which security policy defined at the infrastructure layer extends automatically to AI workloads, application containers and operational tooling.

SECURE
Cisco AI Defense
Model-level protection and policy enforcement across the AI Factory environment — monitoring model-validation events, enforcing access controls, and feeding security telemetry directly into Splunk.
ENFORCE
Isovalent Cilium & Tetragon
eBPF-based container networking and network-policy enforcement across OpenShift, with Tetragon extending visibility to the process level — syscall telemetry, workload identities and runtime behavior.
OPERATE
Cisco UCS, Intersight & NVIDIA
UCS and NVIDIA GPUs form the compute foundation; Cisco Intersight adds lifecycle management and operational telemetry — GPU utilization, infrastructure health and automated lifecycle tracking.
OBSERVE
Splunk Enterprise & Observability
Splunk integrated with OpenShift, Intersight, Isovalent and infrastructure telemetry — GPU activity, model-validation events, packet flows and workload behavior in a single operational view.
The engagement

From stalled project to production, in five phases.

Deploying the environment was the beginning, not the end — the engagement was built to deliver a working foundation fast, then build operational value systematically.

PHASE 01
Discovery & Architecture
Mapped existing infrastructure, team structure and AI workload requirements; defined the target architecture and integration sequence across compute, networking, container and security layers.
PHASE 02
Secure AI Factory Foundation
Deployed UCS, OpenShift, Isovalent Cilium, AI Defense and NVIDIA GPUs — within days — proving enterprise AI workloads can run securely on-premises with full infrastructure and data control.
PHASE 03
Workload Security & Policy
Configured Cilium for network-policy enforcement and Tetragon for process-level eBPF visibility — workload identity controls and behavioral telemetry beyond traditional monitoring.
PHASE 04
Observability Integration
Integrated Splunk with OpenShift, Intersight and infrastructure telemetry; unified dashboards for GPU utilization, model-validation events, packet flows, infrastructure health and workload behavior.
PHASE 05
AI Operations & Lifecycle
Demonstrated AI-assisted troubleshooting and blast-radius reduction by correlating ServiceNow, Intersight, OpenShift and Splunk data; connected lifecycle data to proactive server planning.

How each priority was delivered

In less than a week, the engagement delivered what internal teams had been working toward for months — a validated Secure AI Factory environment with workload-level security and observability already in place.
Outcome of the engagement
The outcome

A validated Secure AI Factory — and a platform to build on.

Validated AI infrastructure

Cisco Secure AI Factory proven on-premises in a complex, regulated enterprise environment — months of stalled deployment turned into a working, validated platform within days, and a credible foundation for the provider’s broader AI strategy.

Enhanced security & control

Isovalent Cilium and Tetragon, combined with Cisco AI Defense, delivered workload-level enforcement and process-level behavioral visibility across AI and application workloads — capabilities traditional network tooling could not provide.

Unified observability & operations

Splunk across OpenShift, Intersight and infrastructure telemetry gave operations a single view of GPU utilization, model-validation events, infrastructure health and workload behavior — enabling faster troubleshooting, proactive lifecycle planning and AI-assisted incident response.

A stronger Cisco platform

The engagement validated Cisco’s Secure AI Factory architecture in one of the most demanding enterprise environments in healthcare, expanding the provider’s footprint across UCS, Intersight, AI Defense and Isovalent — and reinforcing Cisco as its strategic AI-infrastructure standard.

Questions

About this use case.

Who was the customer?

We do not name customers without explicit written permission. This engagement was delivered for a major U.S. healthcare provider with an enterprise-scale server estate and on-premises AI infrastructure spanning compute, GPU, container and security layers.

Is this a delivered engagement or a proof of concept?

Delivered. Ngenium engineers deployed and validated the Secure AI Factory foundation in the provider’s environment — this is a completed engagement, not a lab exercise or a proof of concept on synthetic data.

Which technologies were involved?

Cisco Secure AI Factory: Cisco UCS with NVIDIA GPUs and Cisco Intersight for compute and lifecycle; Red Hat OpenShift with Isovalent Cilium and Tetragon for container networking and process-level security; Cisco AI Defense for model-level protection; and Splunk Enterprise & Observability as the unified observability layer.

Have a stalled AI infrastructure initiative?

We provide solutions. We deliver success. Bring Ngenium into your next Cisco AI infrastructure opportunity.

Schedule a Scoping Call