Everyone has an opinion. Nobody has evidence. In two weeks I get you a straight answer: fixed scope, fixed price.
At Experian, product manager for ML in high-risk credit decisioning systems (ModelOps)At Progress, product lead of the AI agent observability platform (GA 2026)#6 Product of the Day on Product HuntOpen-source agent-skills toolkit on GitHubWebinar host to developer audiences
For teams with an AI agent or LLM feature in production who can't confidently answer “is it working?”
For companies of 10-100 people that know AI should be saving them time but have no plan - or a shadow-ChatGPT problem.
30-minute call fixed quote start within 2 weeks
Hi, I'm Lyubo. By day I'm the product lead of Progress AI Observability, a platform for making AI agents trustworthy in production. On the side I'm the founder of Babuger, an AI SDR platform that runs outreach end-to-end. Before that I was a product manager at Experian, running ML in high-risk credit decisioning systems (ModelOps), where a model you can't explain or monitor doesn't ship. Which means I sit on every side of your problem: I've run regulated ML, I build the tooling that catches AI failures, and I run AI agents against real customers with my own money on the line.
You work with me directly. No handoff, no juniors, no account manager.
I work on this problem in public. The essays and free tools below are the same thinking you'd be buying, so judge it before you book.
A field guide to the five ways agents misbehave in production, and how a trace turns each one from a mystery into a span you can see.
Read the field notes →Long-form explainer on why agents fail in production and what a useful stack actually covers.
Read the explainer →Free generator for LLM-as-Judge evaluator prompts. Paste a trace, get a research-grounded prompt with a built-in stress test.
Free analyzer for MCP servers and tool-call schemas. Static + LLM-assisted findings cited to primary research, plus a portable test pack.
Occasional field notes on making AI agents provably work in production. No pitch, unsubscribe anytime.
Double opt-in via Buttondown.
Because you're buying an outcome, not my hours.
Yes. On-site available in Europe.
Two openings per month. After the fit call you get a fixed quote, and we start within 2 weeks.
Teams wanting a body-shop developer, or enterprises that need a vendor with a procurement portal.
If the fit isn't there, you'll know in 30 minutes.