← Home Work with me

AI agents that hold up in production

Four ways to work together. Each one links the proof already on this site. The first step is always a 30-minute call: no price list, because every team starts from a different place.

AI agent reliability audit

Who it is for
Your agents work in the demo and break in production: loops, failed tool calls, answers nobody can verify.
What you get
  • A review of the harness: tool contracts, context, error recovery, guardrails.
  • An eval set built from your real failures, so the next prompt, model or harness change shows whether agents got better or worse.
  • A written plan: fixes in order of how much production risk each one removes.

Talks and workshops

Who it is for
Meetups, conferences and internal team sessions on AI agents in production.
What you get
  • Topics: verification as the new bottleneck, red test first with agents, real-time voice AI.
  • A hands-on workshop on your own repo, or a talk built on the receipts from this site.
  • In English or Russian.

Tell me the problem on a call. I reply within 24h.

On a laptop? Scan with your phone to save my contact or book from there.
COPIED TO CLIPBOARD