About

Hi, I’m Andrey. I design cloud platforms and AI systems, and I still build them with my own hands. This site holds both halves of that life: a workshop for the engineering, and a quiet nook for whatever else I think about.
The short version
- Cloud and AI architect with 25+ years in IT.
- Azure first, AWS when the job calls for it, Terraform everywhere.
- Recent work: AI agents, document processing with LLMs, real-time data pipelines.
- Microsoft Certified: Multi-Agent AI Solutions Expert (AI-500, November 2026) and Azure AI Apps and Agents Developer Associate (AI-103, October 2026).
- Earlier AWS certifications: Solutions Architect and SysOps Administrator (2019).
- Based in New Jersey, working remotely with US companies.
What I do
Cloud architecture. Azure PaaS platforms, hub-and-spoke networking, private connectivity, identity and governance. The parts nobody notices until they break.
AI and agents. Azure AI Foundry, Azure OpenAI, AI Search, Amazon Bedrock, MCP. Agents that answer questions over real enterprise data, and document pipelines that check what the model says before anyone trusts it.
Delivery automation. Terraform, GitHub Actions, Azure DevOps. Pipelines with gates that people do not need to work around.
Data platforms. Change data capture and streaming, Data Factory, Synapse, data lakes.
Troubleshooting. Any application or service, at any layer, read through its logs, traces and metrics. It comes naturally to me, and my favorite moment is when “it doesn’t work” turns into one wrong line.
How I got here
I studied physics (quantum field theory) at Tomsk State University. It left me with one habit: I do not trust a result I cannot reproduce.
Then came about fourteen years of building security and networking products: head of QA, head of R&D, and finally CTO. After that, six years as an Azure architect on a platform that grew to 10+ regions and 100+ workloads. Since 2025 I have worked as a consultant, most recently on AI agents and data platforms for clients in financial services and retail.
Every step added the same rule. Design it, then prove it works by building it.
Selected work
- A data agent that speaks plain English. A natural-language query agent over enterprise data, built with Amazon Bedrock AgentCore, MCP, hybrid vector search and an ontology.
- Document processing with LLMs. Four services for OCR, classification and validation. Prompt hardening and quote verification made the validation step more accurate.
- Real-time change data capture. A pipeline on AWS DMS, Kinesis and Glue streaming, with infrastructure as code and latency monitoring.
- Azure Government from scratch. A federal government environment: tenant, subscriptions, landing zones, networking, PaaS services, VPN and ExpressRoute to the client’s data center, identity and governance.
- Azure at scale. Environments across 10+ regions for 100+ workloads. Reusable Terraform modules cut deployment time by 60%, and rightsizing and automation cut cloud spend by 30%.
- CI/CD for commerce. A GitHub Actions platform for Salesforce Commerce Cloud storefronts, with branching and promotion workflows for four parallel teams. Read the write-up.
- Open source. AI Collab Hub, a shared workspace for several AI assistants, and an IaC MCP server that keeps AI-generated Terraform inside approved modules. More on the GitHub page.
Why “hands-on”
Most of what I write here comes from something I built, broke, or had to debug at an inconvenient hour. The Labs let you break the same things yourself, in the browser, with nothing at stake.
The other half
When I am not working I walk, and thoughts tend to show up on the walk. Some of them end up in Thoughts: short notes, a comic about an engineer and his AI agent, and “Human and AI”, a slow series of essays about what we mean when we use human words for machines. That corner is also available in Russian, my first language.
What I am busy with this month is on the Now page.
Get in touch
LinkedIn, GitHub, or the contact form. I am open to new roles and projects in cloud and AI architecture, and I am always open to a conversation. Working together says how that usually starts, and my resume is on this site.