About Us
Building the reliability layer for production AI agents
Why we built Blue Guardrails
Agents are moving from answering questions to making decisions, using tools, and executing workflows. As their autonomy grows, teams need a way to understand, evaluate, and steer their behavior in production.
Blue Guardrails gives product and engineering teams one reliability layer to detect issues, compare configurations on production traces, steer agents while they run, and continuously improve them.
We are building the infrastructure that lets teams deploy more capable agents without giving up control.
Our Team
Miriam brings deep technical expertise in large language models and years of experience building production AI systems across legal publishing, news media, finance, and government.
At deepset, she led forward-deployed engineering teams and helped enterprise and public-sector organizations take semantic search, RAG applications, and agents from product design to production.
Mathis is a former AI engineering lead and VP of Product at deepset, where he built and scaled its enterprise AI platform from the ground up.
He has published research on large language models, won a Kaggle competition on AI-based text complexity assessment, and speaks at international AI conferences.
Blue Guardrails is built in Berlin and available with EU data residency, private cloud, and on-premises deployment.
Go deeper
Ready to make your agents reliable in production?