Oren Cohen founded Converra to automate the work of keeping production AI agents reliable: diagnosing failures, generating and testing fixes, and verifying on real traffic whether each fix worked.
He previously founded Buildup, a construction technology company later acquired by Stanley Black & Decker. As VP Product Growth at Totango, he owned AI end-to-end, from design through production, while re-accelerating ARR growth from about 10% to 43% year over year.
He studied Computer Science at the Technion and is based in New York City.
Articles
- Closed-to-open model transfer for production agents
Series · 48 parts
- How to Verify an AI Agent Fix After Deployment
- How to Turn Production Agent Traces Into Improvements
- AI Agent Incident Postmortem Template: What to Record After a Failure
- How to Build an AI Agent Evaluation Harness That Survives Production
- Context Engineering After Launch: An Operations Guide
- How to Build an AI Agent Regression Suite From Production Traces
- AI Agent Reliability Is Not One Score: A Practical Scorecard
- How Many Simulated Conversations Do You Need? Stop Looking for a Magic Number.
- Your LLM Evals Passed. Why Did the Agent Still Fail in Production?
- How to Migrate a Production AI Agent to GPT-5.6
- Without Production Verification, an AI Agent Improvement Loop Is Still a Hypothesis
- Why AI Companies Keep Humans in the Loop
- Your LLM Judge Is Not Ground Truth
- How Converra Evaluates AI Agent Performance
- Best Tools to Fix Production AI Agents (2026)
- Best AI Agent Evaluation Tools (2026)
- Best AI Agent Observability Tools (2026)
- Why AI Agents Regress After Prompt or Model Changes