Edra solves a specific, expensive problem: general-purpose AI agents deployed inside enterprises start from zero, with no knowledge of how that company actually operates. Founders Eugen Alpeza and Yannis Karamanlakis, both ex-Palantir, built Edra to fix this. Yannis created the Forward Deployed AI Engineer role at Palantir and led an AI recruiting search engine that raised placement rates 129%. Eugen built Palantir's U.S. commercial motion and launched its AI Platform under CEO Alex Karp. They have known each other for 13 years.
The technical approach is worth reading closely. Edra does not ask humans to write documentation. It ingests data the company already produces: support tickets, emails, logs, chat histories. From that, it builds a living knowledge base that reflects real operational behavior, not idealized process maps. The system updates itself as it is used, and it is transparent: every inference is visible and editable. First production deployments are in IT service management and customer technical support, where structured data is plentiful and agent automation is proving out fast.
Sequoia's Luciana Lixandru led the investment. The piece is short but worth reading in full for one specific reason: the argument about why black-box fine-tuning fails at enterprise scale, and why transparency in the knowledge layer is the actual unlock for agentic automation. That argument has implications well beyond Edra's current use cases.
[READ ORIGINAL →]