I architect the platform layer
AI products run on.
Cloud-native SaaS on Azure — and, on my own time, on bare Kubernetes I manage myself.
Fourteen years designing and shipping enterprise-scale SaaS across finance, audit, and tax — AKS, Istio, APIM, Terraform, full observability. Lately: shipping AI products end-to-end and running the infrastructure under them without the managed abstractions, so I understand why it works, not just that it works.
The stack, all the way down
SmartDelivery — five .NET microservices I run on a self-managed k3s cluster with an Istio mTLS mesh. This is the live topology, not a stock diagram. Every node maps to a manifest in the public infra repo.
Anyone can deploy to AKS. I stripped away the managed layer and ran it on bare k3s to learn what AKS quietly handles — overlay networking, kubelet cert trust chains, CoreDNS across control-plane restarts, API-server advertise addresses. The distance between "it works on AKS" and "I understand why it works" is real. This closed it.
Verifiable work
Solo, full-stack, and public. Click through — the repos and live products back every claim.
SmartDelivery
Five microservices with MediatR/CQRS, JWT, FluentValidation, Polly resilience, and Serilog→Elasticsearch with cross-service correlation IDs — deployed on a k3s cluster I run myself, behind an Istio mTLS mesh, with GitHub Actions CI/CD. Documented down to an infra spec, observability architecture, and a repo-vs-cluster audit with migration runbook.
Doodlecast
Turns any blog-post URL into an illustrated, animated explainer with voiceover. Built and shipped solo — an event-driven pipeline of seven queue workers behind a Fastify API, progress streamed to the browser over SSE. Originally fronted by APIM; I took it out deliberately and moved rate limiting to the edge, so the whole thing now runs on free tiers and scale-to-zero with a single ~$5/mo registry as the only fixed cost.
NoConfusion
An AI decision-making app with multi-agent group chat and debate modes, custom agent personas, tool use (web search), context-window management, and a credit-based usage economy — the product-side counterpart to the infrastructure work.
Track record
Fourteen years, one clean upward line — from writing the code to owning the architecture and leading the teams.
Led .NET/Azure architecture for a cloud-native audit SaaS on AKS — HPA/KEDA autoscaling, Istio, Terraform, enterprise observability.
Architected .NET Core + Angular microservices for a tax & finance auditing platform on Azure PaaS.
Designed secure identity lifecycle & auth systems; scalable .NET Core backends on Cosmos DB and Functions.
Built a global bike-rental SaaS (15 cities, 8 countries) on .NET Core + Azure PaaS; governance & compliance automation.
Certified AZ-305 (Azure Infrastructure Architect) and AZ-400 (DevOps Solutions). Udacity — Intro to ML with PyTorch.