About
Frederik Rybansky
I am an AI infrastructure and agent engineer in Bratislava. I build the platform layer behind AI products — retrieval, agents, evaluation, observability — and the customer-facing assistants that sit on top of it. I have been doing software professionally for around a decade, and the last few years almost entirely on AI systems.
My path into AI was indirect and probably useful. I started in enterprise software, spent years on multi-tenant platforms and integrations at BSS PRAHA, then on a public content API at Audioline and IoT telemetry at BatteryCheck. All of it taught me the thing that AI projects need most and rarely have: production discipline.
A model that is wrong in a demo is a bug report. A model that is wrong inside an ERP workflow is an incident. I care about the second category — about systems that are observable, testable, permissioned and reversible — because that is where the actual value of AI in a company is decided.
Today I work with a small number of clients at a time, usually as an extension of their engineering team. I write production TypeScript and Python, I design the architecture, I review your team's pull requests, and I make sure the thing still works after I leave.
What I do
Mostly AI infrastructure and enterprise AI agents, plus AI chatbots on websites. Also ordinary software — APIs, integrations, internal tools — because AI projects still need a database and somebody who knows how to build one. The full list is on the services page.
How I work
Written plan before code, working software every week, and measurement agreed in week one. I prefer boring technology in the plumbing and interesting research in the product. I do not run discovery that quietly becomes an invoice.
Languages
English, Czech and Slovak. I work with clients in all three, and every site and system I build is designed for all three rather than translated afterwards.
Where I am
Bratislava, Slovakia, CET. Remote across the EU and beyond, structured around async written updates rather than meetings that would happen at 3am for somebody.
Trajectory
- 2026 Independent AI infrastructure & agent practice Working full-time on enterprise AI platforms, agent systems and multilingual website assistants. Mostly regulated environments where the answers have to be defensible.
- 2025 Deep shift to applied AI Shipped retrieval infrastructure with hybrid search and reranking, built evaluation harnesses that run in CI, and moved client work from pilots to production.
- 2024 Enterprise integrations at scale Typed API platforms and integration layers across enterprise customers, with the reliability and observability work that made them maintainable.
- 2023 BatteryCheck shipped Battery diagnostics running at fleet scale: device management, telemetry ingestion, threshold alerting and workshop reporting.
- 2020–2022 Audioline content platform and public API Rebuilt a CMS and its public API around typed contracts, caching and predictable latency, with a versioning policy clients could rely on.
- 2016–2020 BSS PRAHA — enterprise platform engineering Multi-tenant services, customer integrations and data pipelines inside a Czech enterprise software company, working directly with customer engineering teams.