Clean Evals Book a working session

Evals as a service

We build and maintain evals for your AI agents.

Book a working session

Teams we’ve helped

How we work

We take care of your evals.

We turn real user requests into test cases, agree on success criteria with your team, and keep evals up to date as your product and user behavior change. You see where the agent fails and whether your changes make it better.

In your infrastructure.

All data stays in your infrastructure. You get all the code and evals.

Audit

1 week

$1,500

An eval roadmap for your agent.

  • Priority workflows and test cases, including known failures.
  • Success criteria for each workflow and a list of missing data or context.
  • An implementation plan covering test coverage, setup, and acceptance criteria.

About 2 hours with your product expert.

Implementation

8 weeks

$20,000

Evals your team can use independently.

  • Infrastructure to execute evals, capture traces, and produce reports.
  • Hill-climb evals and regression evals for the agreed workflows.
  • LLM graders calibrated against expert reviews and validated on separate cases.
  • Baseline reports with failure examples and priorities for improvement.

2–3 hours a week with your product expert.

Continuous improvement

Monthly

$5,000 / month

A dedicated eval engineer.

  • Updated eval cases and success criteria based on production traces.
  • Grader recalibration when needed.
  • Monthly reports on performance, coverage gaps, and completed work.

A weekly 30-minute check-in with your team.

The team

Meet the founders.

Max Petriev

Max Petriev

Co-founder

LinkedIn

Max has spent a decade building software and leading engineering teams. He co-founded TheStage AI as CTO and built backend infrastructure at VK and Delivery Hero. His background includes a mathematics degree from Moscow State University and PhD research in statistics.

Eli Polterovich

Eli Polterovich

Co-founder

LinkedIn

Eli’s background spans startup operations and venture capital. He co-founded Muni, which builds AI tools for municipalities, and worked in operations at HoloRay. He studied at McGill University, where he was an analyst with McGill Ventures, the university’s student-run venture capital group.

Want to know what your agent does well—and where it needs work?

Book a working session