IT Leadership · AI Strategy · Enterprise Architecture

Andrew Katana

29 Years in IT | Self-Hosted Expertise — Building robust, scalable systems for Enterprise Data, Agentic Workflows, and Enterprise Tech strategy.

(@katana) · atkatana.com

Core Focus Areas

Retrieval Augmented Generation (RAG)

Production retrieval pipelines — hybrid search, reranking, and vector persistence tuned for real workloads.

Agentic Workflows & Process Automation

Stateful, event-driven agents that extend teams — automated with safety-first design.

Infrastructure Architecture Strategy

From bare metal to multi-cloud — repeatable platforms teams can depend on.

Identity, Access & Security

Bridging the gap between robust governance and high-velocity innovation.

About Me

Ben Bulben, County Sligo, Ireland
Ben Bulben, Co. Sligo, Ireland — one of the reasons the site isn't all dark grids and code.

I'm Andrew Katana — an IT leader who builds the platforms teams depend on. Over 25 years, I've founded a Cloud Center of Excellence, scaled a SaaS portfolio from $240M to $450M in revenue, built a $26M professional services division, and led 75+ engineers across four global regions through cloud migrations that moved billions in software from on-premise to AWS.

Today, I design and deploy production infrastructure — container orchestration, IaC automation, AI-powered operations, and disaster recovery — across distributed, multi-region environments. I run a three-machine self-hosted AI stack with agentic workflows, RAG pipelines, and MCP-protocol agent orchestration. I don't just advise on transformation — I do it, hands-on, every day.

My strength is translating emerging technologies into practical, scalable solutions that deliver measurable business value — balancing innovation with risk, security, and compliance.

Ask AI Say hello →

Ask AI — Query my self-hosted AI on any topic. Trained on my real work, not generic answers.

Selected Projects

Cloud Center of Excellence — Micro Focus / OpenText

Problem: 10+ product groups with inconsistent cloud adoption, no shared services, no governance. Each team built their own AWS accounts with ad-hoc security.

Built: Founded the CCoE — deployed Landing Zone across 47 AWS accounts, hardened AMIs (Linux/Windows), account factory automation, Vault backup, self-service portal, FinOps reporting, and 50+ security guardrails. Managed 3 SRE squads across 4 regions through biweekly sprints.

Result: Standardized cloud adoption across the entire SaaS portfolio. Reduced operational overhead while improving reliability. Scaled the platform supporting $240M→$450M in annual revenue.

AWS Organizations · Control Tower · Terraform · IAM · FinOps

Project CT — Data Center to Fargate

Problem: Multi-billion dollar SaaS portfolio running on-premise VMware. Each product group had different migration readiness, and no reference architecture existed.

Built: Designed the reference architecture migration from on-prem to ECS Fargate with GitOps CI/CD. Led the "Lift and Improve" pattern — not just moving servers, but enhancing HA and DR for global clients. Embedded DevSecOps with automated vulnerability gates.

Result: Project CT became the standard pattern adopted by Stack A and B product groups, defining the blueprint for all subsequent cloud migrations across the portfolio.

AWS ECS · Fargate · GitOps · CI/CD · CloudFormation

Self-Hosted AI Stack — Production Agent Infrastructure

Problem: Enterprise AI tools are expensive, opaque, and don't integrate with real workflows. Needed a production-grade AI infrastructure that's transparent, extensible, and runs on commodity hardware.

Built: Three-machine production pipeline: Ollama (GPU inference on RX 6900 XT), Qdrant (hybrid dense+sparse vector search), Playwright MCP (browser automation), and specialized agents for job search, knowledge management, and content creation. 387+ knowledge entries, MCP protocol for agent-to-tool communication, cron-driven automation.

Result: A fully self-hosted AI platform that handles RAG, browser automation, email triage, job screening, and content deployment — all running locally with zero cloud AI dependencies.

Ollama · Qdrant · RAG · MCP · Docker · Playwright

IAM Redesign — Enterprise Identity Federation

Problem: 35 AWS accounts with fragmented identity — each team managing their own IAM, no SSO, no centralized governance. Security audits flagged inconsistencies across the board.

Built: Led Cloud Services Acceleration IAM redesign: Okta SSO federation, AWS Identity Center, Account Factory for Terraform, centralized IAM with SCPs. Integrated lessons from Control Tower deployment into a repeatable pattern.

Result: Unified identity across 35 accounts. Single sign-on for all engineers, centralized access governance, and automated account provisioning with security guardrails baked in from day one.

Okta · AWS Identity Center · Terraform · Control Tower

What I Build

Enterprise AI Architecture

Designing production-ready AI systems — retrieval pipelines, vector persistence, and LLM orchestration tuned for enterprise scale, security, and governance.

  • • Production RAG (hybrid retrieval, reranking)
  • • Vector DB design & tuning (Milvus, Weaviate, Qdrant)
  • • LLM evaluation, guardrails & MLOps

Infrastructure & DevOps

Architecting hybrid and multi-cloud platforms with repeatable automation — from bare metal and self-hosted stacks to managed cloud services.

  • • Multi-cloud architecture (AWS, Azure, GCP)
  • • Kubernetes & platform engineering
  • • Infrastructure as Code (Terraform/Ansible)

Agentic Automation

Building stateful, event-driven agents that extend teams — lead gen, ticket routing, and automated content workflows with safety-first design.

  • • LangGraph / AutoGen orchestration
  • • Event-driven triggers (Webhooks/Redis)
  • • Prompt-engineered safety guardrails

Teams & Team Building

Everyone wants experienced SREs and cloud engineers. Nobody wants to build them. I build them — turning inherited engineers into architects and SREs through a deliberate people pipeline.

  • • Grew a 2-person team into 3 SRE squads across 4 regions
  • • Promoted engineers to Principal SRE / Solution Architect
  • • "You can buy tools and rent accounts, but you can't buy teams"