Backend, infrastructure, and the unglamorous plumbing that keeps AI products alive in production.
ALL CLIENT PROJECTS DELIVERED THROUGH A POLISH SOFTWARE HOUSE, 2021 - PRESENT.
A private AI platform for enterprises whose data can't be sent to a public model. I own the architecture end to end - retrieval, the agent runtime, and the infrastructure underneath it. Our own MCP servers give the agents around 40 data connectors - Gmail, Jira, Confluence, Slack and the rest of the systems a company actually runs on; they delegate to subagents and keep long-lived memory; anything they generate runs inside sandboxed, air-gapped runtimes on gVisor and Kata Containers. It ships as a Kubernetes deployment per customer - their cloud or fully on-prem, GPUs included.
Python · Kubernetes · RAG · MCP · gVisor / Kata · On-prem GPUA GitOps platform for teams running Kubernetes in more places than they planned to. I led the backend orchestrating the agents installed on customer clusters and owned the infrastructure around it - CI/CD, secrets management and deployments across AWS, GCP and Azure. I also gave every engineer a disposable backend of their own, deployable on demand from GitHub Actions, which quietly ended the queue for shared staging.
Node.js · Kubernetes · GitHub Actions · Terraform · AWS · GCP · AzureInternal event platform for conferences of several thousand attendees - check-in, agendas and live updates. A short stint between two longer projects: backend work, and keeping the system standing on the days it mattered.
Node.js · GraphQL · MongoDBA video learning platform for executives and managers - an internal Vimeo with courses, roles and reporting wrapped around it. I worked on the backend and, long before it was fashionable, wired an early OpenAI model into it to categorise courses and generate tags automatically.
Node.js · GraphQL · Elasticsearch · OpenAI APIPlatform for a fast-growing US property management company with thousands of units under management. Backend work, including the communication layer the product ran on: VoIP calls and support mailboxes handled inside the product instead of in somebody's personal inbox.
Node.js · Express · MongoDB · VoIPBuilt, deployed and maintained websites for small businesses and individuals. Where I learned that somebody has to keep a thing running long after it launches.
WordPress · JavaScript · Hosting / DNSBuilding backend systems and AI infrastructure - mostly retrieval, agents, and the unglamorous parts that decide whether either survives contact with production.
Five years, one employer, five industries: property, e-learning, events, developer platforms, enterprise AI. Since 2021 that employer has been Acaisoft, a software house in Gdańsk. The domain keeps changing; the job of making something scale and stay up doesn't.
On the side I'm building two things - somewhere to collect rides and hikes, with GPX and FIT files, photos and maps, and a tool for analysing and picking stock portfolios.
M.SC. BIG DATA ANALYSIS, UNIVERSITY OF GDAŃSK, 2026 · B.SC. INFORMATICS & ECONOMETRICS · CHANCELLOR'S SCHOLARSHIP ×3