We started Jean Technologies to solve a problem we kept seeing across industries: products whose whole value depends on understanding a person, built on infrastructure designed to understand documents.
The embedding models and retrieval systems that work well for finding relevant documents do not work well for representing people. The objectives are different. The data is different. The evaluation criteria are different.
What we are building
Jean Technologies trains foundation models of human behavior. We learn representations of people from outcome data, what someone actually did rather than what they said about themselves, and deploy them as API-accessible services.
The reason to build at the representation layer is that the applications people ask us for are all the same problem wearing different clothes. Memory, matching, recommendation, personalization, and simulation are downstream tasks. Each one is a question you can ask a sufficiently good model of a person, and none of them is solvable if that model is missing.
Two pieces of that stack are available today. Person Search exposes the representation directly for retrieval over people. The Embedding Adapter enables zero-downtime migration and cross-silo federation between vector databases.
Who we work with
We work with platforms where understanding the user is the business: recruiting and hiring, dating and social, marketplaces, and investor and founder connectivity. These are the domains where outcomes are recorded cleanly enough to train on, which is also what makes them the right places to start.
How we engage
We provide our technology through two models: platform access via API for teams with mature ML infrastructure, and forward-deployed engineering engagements where we embed with your team on a fixed-outcome basis.
We are a small, focused team based in Palo Alto. If you are building a product that has to understand the person using it, we would like to talk.