Hello, l13v

I'm Lyuboslav Petrov — an AI solutions architect and full-stack researcher based in Sofia. l13v is the short form of lyuboslavpetrov, and this site is the home for my work: a decade-plus of Python engineering and architecture, currently pointed at the question of how AI agents and human engineers actually build software together.

Most of my time goes into two connected threads. The first is deterministic data-access planes for AI agents: systems that introspect the databases an operator already runs, recover an ontology from the data's own structure, and project complete, predictable query surfaces from it — so an agent's access to data is exact rather than guessed. The second is agent-augmented engineering practice: harnessing frameworks, development methodologies for human–agent collaboration, standards enforcement through agents, and memory that survives long-running sessions.

This blog will cover the engineering behind both. Expect posts on agent and data-plane architecture, evaluation discipline — how to know whether an agent-facing system actually works, beyond a passing test suite — and ontology recovery from heterogeneous backends. Occasionally something from the longer arc: computer vision for sports analytics, industrial control systems, or whatever else is on the bench.

The writing here aims to be concrete: real designs, real measurements, real failure modes. If any of it is useful to you, or you want to talk about applying it, get in touch.