Skip to content
About Orq AI

We built the tool we needed and couldn't find

After shipping LLM-powered features that broke silently in production, we decided the right answer was an observability layer built for the shape of agent work.

Why we started Orq

In 2023 we were building an LLM-based document processing feature at a previous company. It worked perfectly in staging. In production, the grounding started failing on a specific document format we hadn't tested. We didn't know until a customer told us, three weeks after the regression was introduced.

The problem wasn't the model. It was that we had no trace of what the agent actually did. No score on the output quality. Nothing blocking the deploy when the prompt change landed. We fixed it by adding logging, but that was reactive. We needed something that caught the regression before the deploy.

We looked for existing tooling. Distributed tracing systems weren't built for the LLM call shape. Eval frameworks ran offline in notebooks. Nothing connected tracing to evals to CI in one place. So we built it.

Orq AI launched in 2023 from Rotterdam. We are a small team with a sharp focus: make it possible for any engineering team to ship AI agent changes with the same confidence they ship backend API changes.

The team

The people building Orq

Sohrab Hosseini

Sohrab Hosseini

CEO and Co-founder

Sohrab spent his career building production ML infrastructure before starting Orq. He shipped LLM systems to production and grew frustrated by the gap between offline eval notebooks and real agent behaviour under live traffic.

Fenna van Dijk

Fenna van Dijk

Head of Product

Fenna built developer tooling at two previous startups and cares deeply about the gap between what engineers need and what monitoring products actually give them. She leads product strategy and design at Orq.

Jonas Park

Jonas Park

Lead Engineer

Jonas built the ingestion pipeline and release gating system that powers Orq. He came to the team with a background in distributed systems and low-latency data pipelines, and a specific interest in making evaluation systems actually run at the speed of CI.

Based in Rotterdam

Blaak 40, Floor 8, Rotterdam 3011 AA, Netherlands. We are a small team building Orq full-time from Rotterdam. If you want to join or talk to us: [email protected]

Ship agents with confidence

Get tracing, evals, and release gating running on your agent stack today. Free to start.