I'm Dmitrii Lutskov, an architect for complex, data-heavy systems: high-load e-commerce, geospatial platforms, network infrastructure, database integrations. I plan, architect, and direct delivery, and I'm drawn to any problem hard enough to be worth taking apart. Project, advisory, and consulting work.
I decide how a system should be built, then direct AI agents to execute that design. I use the best tool for each task. If designing or delegating gets a better result faster, I do that. When the agents fall short, I go deep myself.
That last part is not theory. On a hard Kubernetes ingress problem the models kept producing nonsense, so I took it apart by hand, used targeted queries only where they actually helped, and mapped, understood, and fixed it in three days. The fundamentals are still there. The AI is a lever, not a crutch.
The real bottleneck in good software was never typing. It is judgment: which design survives scale, which shortcut becomes next year's outage. That is what three decades of building gives you, and it is what a model does not have on its own.
These are the areas where I go deepest. They are examples, not limits: I take on hard systems of many kinds.
Since 2009 I have owned the technical side of an online store with 60,000+ SKUs. Codebase, automated pricing and supply chain, infrastructure, and the marketing analytics on top.
I came into GIS on the vendor side, building the commercial platform itself rather than applications on top of it, so I know the machinery and not just the APIs. I now build on the open-source stack: portals, routing, tiling, and the spatial databases underneath. I keep working with FME and data integration.
Monitoring systems, client-server platforms, and DNS infrastructure, including a full-featured DNS service with DDoS and bot protection.
Full-stack systems and APIs since the early web. Backends, payment and platform integrations, and the database work underneath.
Five threads. The oldest opened in 1990, the newest in 2024, and not one of them was ever closed.
Low-level systems, CAD and enterprise client-server, in C/C++ and Oracle, when hardware was tight and every byte and cycle had to behave. That is why I can still see a system end to end.
Linux servers and mail from 2000, Nginx from 2004, Kubernetes since 2018, with the log pipelines and Go tooling around them. The part most developers hand off, I own.
Started on the vendor side, building the commercial GIS platform itself rather than applications on top of it. Which is why I read spatial systems from the engine rather than the interface.
Owning a 60,000-SKU store: Rails, automation, infrastructure and the marketing analytics behind it. So engineering serves revenue instead of sitting beside it.
Directing AI systems to build, with the judgment to keep what they produce production-grade. The newest thread, and the only one the other four are holding up.
A timeboxed review, about three days deep in the repository and logs. You get a written report: the critical risks and a step-by-step stabilization roadmap. This includes rescuing AI-generated codebases that ran in a demo and broke under real load.
Ongoing technical direction for a founder or team. Architectural decisions, infrastructure, and development process, without a full-time hire.
A system designed and delivered end to end, with AI-directed execution, when you need the thing built and not only advised.
Project, advisory and consulting only, not staff roles. I work primarily async: written specs, code, and recorded walkthroughs, with calls as needed.
Anonymized engineering case studies. The architecture, stack, and engineering are real.
One of the longest-running online electronics retailers in its market, with a Rails core, a C# ERP integration, and automated pricing across more than a dozen supplier feeds. About 60,000 SKUs, in continuous production since 2009, run by one engineer.
Built end to end, from a hand-written C socket listener that speaks the device protocol, through PostGIS and a real-time backend, up to the Flutter apps and a live map.
A layered defense in front of customer sites: authoritative DNS, an external layer for L4 and volumetric attacks, and an internal senginx protector doing L7 bot mitigation, whitelists, and port forwarding.
Took over a raw e-commerce platform after the previous team left and reworked it over three years, from falling over near a thousand users to tens of thousands of concurrent live users, on Kubernetes.
"He consistently delivers significant results with minimal supervision, often succeeding where others would require considerably more support. Reliable, straightforward to work with, and committed to high professional standards."
Business development manager · enterprise GIS · ~20-year collaboration
"Dmitrii architected and led the engineering of our DNS and traffic-protection platform. He owned the technical side end to end: the architecture, the build, and keeping it running in production for years."
Co-founder & CEO · DNS and traffic-protection platform
"I brought the business and the clients; he built what was needed and kept the infrastructure running. Across all those years and projects, he never allowed the technical side to become my problem."
Client · over a decade of mapping and geospatial projects
"He owned the Rails codebase and set the technical direction, and the system stayed stable at a scale where most setups start to crack."
Founder & CEO · 60,000-SKU e-commerce platform
All four are published recommendations on my LinkedIn profile, under the names of the people who wrote them. Clients stay anonymous here; the people do not.
Tell me what you're building, or what's breaking. I reply by email, usually within a day.