Email WitchOps PROFESSIONAL PROJECT / ENTERPRISE AI INFRASTRUCTURE
Extending established Azure infrastructure for the Levi Strauss & Co. AI Super Agent initiative.
01 / THE INFRASTRUCTURE CHALLENGE
An architecture diagram can make a new platform look straightforward. The actual environment has existing workloads, shared modules, governance policies and decisions that other teams already depend on.
That was the setting for this project. The Super Agent initiative aimed to provide a conversational interface that could orchestrate specialized AI agents and connect them to enterprise systems.
My responsibility was the infrastructure supporting those capabilities. I worked on extending the established Azure environment, rather than building the AI orchestration engine.
New services had to fit into the organization’s existing Terraform and Terragrunt ecosystem. Supporting the new requirements could not come at the expense of configurations already using the shared modules.
02 / THE ENGINEERING DECISIONS
I extended existing infrastructure modules to support newer capabilities while preserving backward compatibility for their existing consumers.
That required understanding how the modules were used, their dependencies, and the assumptions existing configurations made about their behavior.
I developed new Terraform modules within the client’s established policies and infrastructure standards. The newer modules also needed to remain compatible with the existing ones.
A reusable module has to fit the environment where teams will maintain it. Introducing a separate automation approach would create another system for those teams to understand.
Creating a resource is only part of the work. Applications still need to reach it, authenticate, and receive the appropriate access within the organization’s security boundaries.
I addressed connectivity and identity requirements within the existing enterprise architecture, considering permissions, dependencies and how the services would work together.
03 / WHAT THIS PROJECT REINFORCED
The interesting challenge was bringing new AI infrastructure into an established platform while preserving the commitments that platform already made to other teams.
This work brought together shared-module upgrades, compatible new modules, governance, networking and identity. They were all parts of the same application infrastructure problem.
A change can be sensible in isolation and still cause problems elsewhere. I approach that by understanding the existing contracts before extending them, and keeping the organization’s maintenance needs in view.
Enterprise AI starts with an organization’s existing infrastructure, its next set of requirements, and the engineering needed to connect the two.
SOURCE / PUBLISHED PROJECT REFLECTION
This case study is adapted from my LinkedIn article, Enterprise AI Doesn't Start with a Blank Slate. It discusses my infrastructure contribution and general engineering principles; proprietary client configurations and internal tooling details are omitted.
APPLY THE APPROACH TO YOUR ENVIRONMENT
Assess Terraform compatibility, identity and networking constraints before extending an established platform.
Separate consulting assessment. The evidence and limitations above remain specific to the original project or demonstration.
Email WitchOps Explore the assessment ↗