From initiative to production-ready platform.
I help startups turn a founder's idea — whether it's a slide deck or a vibe-coded prototype — into a production-grade solution that holds: real distributed-systems architecture, DevOps discipline, and AI applied where it actually earns its keep.
Three lenses, one pattern
Take something fragile — an idea, a legacy product, a startup mid-fundraise — and make it reliable enough to bet a company on.
AI-enabled development
AI woven into both the process and the product: agents and skills driving the work, MCP making what ships reachable by other agents, not just people.
Platform & infrastructure engineering
Architecture that survives funding rounds, multi-tenant load, and infrastructure costs that could sink a startup's runway.
Product development
Open-source products and internal tooling, built to outlive the project they were built for.
Case studies
Six platforms, one thread each: what was broken, what I built, what happened after.
AI-Enabled Development
Mobitru
- Problem
- A mature mobile and cross-browser testing platform, years on the market and inherited from a previous owner, had to stay relevant as AI assistants started doing work engineers used to do by hand.
- Approach
- Drove the platform's shift to an AI-native product: introduced a Mobile MCP Server and a Playwright MCP Server so AI assistants operate test infrastructure directly, authored Claude Code Skills on top, and folded in BrowserKube — my own open-source project — to add desktop browser testing.
- Outcome
- A years-old product repositioned as AI-native testing infrastructure: established platform, open-source extension, and AI-assistant-facing layer, working as one system.
ReportPortal MCP
- What
- Bridge between ReportPortal's test-execution data and AI assistants — natural-language queries over test results from Claude, Copilot, and Cursor.
See also: AIStats — computer-vision video analytics meets the same AI-native approach.
Platform & Infrastructure Engineering
SKAEL
- Problem
- Building a hyper-automation platform — an automation engine plus a RASA-based conversational layer — for enterprise clients, deployable as SaaS or hybrid SaaS+on-prem. RASA itself had no multitenancy story.
- Approach
- Joined pre-product, working directly with the founders (later the company's C-suite) from idea through architecture through build. Built the automation core on Temporal for durable, fault-tolerant workflow execution, and wrote a custom Golang Operator to run isolated RASA instances per tenant.
- Outcome
- "Digital employees" before the term existed — a platform that carried the company through multiple funding rounds.
VSOptima
- What
- Brought an existing platform onto Kubernetes to support both SaaS and on-prem installs, and built a DevOps practice from close to scratch.
AIStats
- What
- Architected a computer-vision sports analytics platform (soccer match video, ML-augmented) with heavy, globally distributed GPU demand. Built the processing pipeline on Temporal and solved GPU cost efficiency with a custom job-allocation scheme across multiple Kubernetes clusters worldwide, scheduled via Kueue.
Product Development
ReportPortal
- Problem
- Test-automation results across the industry had no strong analysis or reporting layer.
- Approach
- Originated the idea, built the first implementation, secured internal funding to get it going, then owned the architecture through a monolith → microservices → Kubernetes-native evolution.
- Outcome
- One of the most widely used test-reporting systems in the industry, with agents for most mainstream test frameworks — 2,000+ stars, 500+ forks, 100+ contributors.
BrowserKube
- What
- A Kubernetes operator that schedules browsers as pods, started as a personal project to fix resource-hungry Selenium and Playwright test runs. It now runs in production as the browser grid behind Mobitru's desktop testing.
Engineer, architect, brevet rider
I'm a software engineer and architect based in Warsaw, Poland — originally from Minsk, Belarus, with a degree in Electronics from the Belarusian National Technical University (BNTU). Over more than a decade in the industry, I've kept circling back to the same kind of work: taking something from "someone's idea" to "a platform that actually runs in production" — usually for startups, almost always with open source somewhere in the foundation.
I like moving fast without cutting corners that come back to bite you later — durable-execution engines instead of duct tape, multitenancy designed in instead of bolted on, infrastructure that's boring in the right ways. Lately that also means bringing AI into the loop properly: not as a demo, but as something a platform is actually built around.
Outside of work: family, a daughter, a dog, and long-distance cycling — brevets, mostly, which is its own lesson in pacing yourself for the long haul.
- Based in
- Warsaw, Poland
- From
- Minsk, Belarus
- Education
- Electronics, Belarusian National Technical University
- In the industry
- 10+ years
- Off the clock
- brevets, family, an Australian shepherd
Always happy to talk platforms, infrastructure, and where AI actually earns its keep.
If you're building something real and any of this looks like your problem, say hello.