Author and engineer
Vladimir Kroz
Reliable AI agents and evaluation, search and information retrieval, and production AI and ML systems and platforms.
Vladimir Kroz is a software engineering professional focused on distributed systems, large-scale data processing, search and information retrieval, machine learning systems, and the infrastructure required to operate reliable AI agents.
At Agenture AI Lab, he writes practical research for engineers building agentic systems. His work connects agent evaluation and multi-agent coordination with lessons from production search and ML platforms: measurable relevance, observable behavior, controlled execution, and infrastructure that remains dependable at scale.
Areas of focus
- Reliable AI agents and evaluations: task design, safety controls, observability, release gates, and production feedback.
- Agentic and multi-agent systems: coordination patterns, failure modes, and architectural tradeoffs.
- Search infrastructure and information retrieval: retrieval, ranking, relevance, recommendations, and large-scale data processing.
- AI and ML systems and platforms: the data, evaluation loops, serving infrastructure, and operational controls behind production machine learning.
Selected articles
- Agent Memory Is a Search System
- The Coordination Tax: When Multi-Agent Systems Are Worth It
- Agentic AI: Towards reliable AI agents
- State of AI Agents in 2025
Profiles
LinkedIn, GitHub, GitLab, Stack Overflow, and Medium. His personal engineering notes are at vkroz.github.io.