Aleksandr Mordvinov
Founder and principal researcher.
More than fifteen years in work where documents carry liability. First on the business side: B2B import and certification of electronics, regulatory permits, and customs clearance with lawyers and brokers, as head of sales and then commercial director. Then marketplace operations and analytics for a portfolio of suppliers.
Since February 2024 he has built the lab’s systems himself: document intelligence, retrieval, speech, agent runtimes, and private deployment.
Education. Bachelor’s degree in finance and credit, Moscow Financial and Economic Institute, 2014. Before that, two years of electronics and computer engineering at the Moscow Institute of Electronic Technology.
2026. EvidenceGene Court at FIND EVIL!, the SANS Institute hackathon for autonomous incident response.
Languages. Russian, English.
How the lab works
The lab is one person and a team of AI agents. Each agent owns one zone: code, research from primary sources, outreach, content, independent review, planning. They share one memory, in which reviewed facts outrank unreviewed notes. Decisions are written down and numbered.
Nothing goes out in the lab’s name — a message, a post, a payment — without the founder’s explicit yes for that action.
The method
- Understand. The problem, the baseline, the cost of an error, and who decides.
- Design. The question, the comparison and the test cases, written before any code.
- Prototype. A working slice on representative inputs.
- Validate. Measured against the baseline, with failures listed.
- Integrate. Into the systems people already use, with a person in the approval path.
Rules the lab keeps
- A number without a source is not published.
- A failed experiment is a result.
- The comparison is always against the strongest ordinary approach.
- A person signs.