Alex Zhdanov
I'm a research engineer working on AI systems, autonomous agents, evaluation, and ML infrastructure.
I like problems where the architecture is still unclear and there is room to experiment rather than simply implement an established solution. Most of my current work is around agent harnesses, benchmark infrastructure, autonomous software-engineering agents, and ways to make complex AI systems easier to evaluate, control, and improve.
I'm currently building opensiro, where I explore infrastructure for harness engineering and cheaper experimentation with AI agents. I also do independent research on generalization, protected evaluation, multi-agent systems, and organizational control.
Outside of individual projects, I'm interested in how increasingly autonomous AI systems should be organized: how they delegate work, recover from failures, verify each other, operate under constraints, and remain understandable to humans.