Private AI / VMware Cloud Foundation / Kubernetes
Notes from the infrastructure side of AI.
I write about private AI platforms, Kubernetes, VCF, NSX, GPU infrastructure, and the operational realities of building AI systems inside enterprise environments.
About
I work where enterprise platforms, networking, and AI infrastructure meet.
Brandon Quantz is a Senior Product Architect specializing in infrastructure, enterprise platforms, virtualization, networking, and cloud infrastructure. His background spans private cloud architecture, software-defined networking, Kubernetes, and operational platform design within enterprise environments.
He is currently focused on learning and building at the intersection of AI infrastructure, hybrid cloud, and modern datacenter operations. He approaches technology from a systems-thinking perspective, with an emphasis on practical experimentation, hands-on lab work, and understanding how emerging technologies fit into real operational environments.
Brandon is a Broadcom Knight for VMware Cloud Foundation and NSX. Through this blog, he shares technical lessons, architecture discussions, lab projects, and real-world troubleshooting experiences from the evolving world of infrastructure and AI.
Outside of technology, he enjoys rock climbing, strength training, and continuous learning. The views here are his own, and technical details are generalized so the writing stays useful without disclosing employer, customer, or confidential information.
Focus areas
The topics I keep coming back to.
Private AI platforms
GPU-enabled infrastructure, local inference patterns, RAG pipelines, workload isolation, data locality, and hybrid model strategies.
VMware Cloud Foundation
VCF architecture, NSX networking, workload domains, platform operations, and how traditional private cloud changes under AI workloads.
Kubernetes and platform engineering
Containerized services, supervisor patterns, registries, scheduling models, and the control planes that make platforms repeatable.
Architecture leadership
Translating technical constraints into decisions: cost, risk, security boundaries, operating models, and what is actually worth building.
Themes
Three threads I am building in public.
Forge
Building a bare-metal-to-platform automation system for repeatable infrastructure rebuilds.
Agentic Operations
Real troubleshooting stories where AI agents and human architecture judgment work through infrastructure failures.
Private AI Foundation
Notes on private AI infrastructure, GPU-enabled platforms, Kubernetes, VCF, NSX, and hybrid operating models.
Latest writing
PAIF Part 3: In-Lab LLMs and the Output-to-Action Firewall
Private AI becomes operational when the model is not trusted to act directly. The useful pattern is an in-lab model, a structured proposal, a policy gate, and an execution path that already has controls.
The Agent Org: Why Safe AI Operations Is a Structure Problem
The safer system is not the agent with the most access. The faster one is not the org with the most agents. Both problems need structure.
Conversational Operations Needs a Platform, Not a Prompt
A useful AI operations layer is not a clever prompt on top of disconnected tools. It needs a source of truth, bounded services, durable telemetry, and a tool surface the agent can actually use.
VCF 9.1 Rebuild: DNS and FQDNs Were the Real Dependency
A failed VCF 9.1 registration path came down to a basic dependency: the names in the bring-up spec did not resolve exactly as the platform expected.
MCP TLS Was Only Half the Security
TLS made the AI tool endpoint safer in transit. It did not answer the harder question: who is allowed to call operational tools in the first place?
The Agent I Already Had
An inbox cleanup story about why agency, execution location, and data boundaries matter more than model intelligence alone.
VCF 9.0 vs 9.1: IP and DNS Planning for Internal Components
A practical planning model for the internal components that need IP addresses, FQDNs, forward DNS, and reverse DNS across VCF 9.0 and VCF 9.1.
Designing PAIF: A Reference Architecture for Tier 3 Private AI
A practical design for private knowledge, local processing, Tier 3 private question answering, Kubernetes-first AI services, and hybrid routing when policy allows.
Contact
Good conversations start with real constraints.
If you are thinking through private AI, VCF, NSX, Kubernetes, or platform architecture, I am always interested in comparing notes.