Soren

Soren is building private AI infrastructure for teams in regulated industries, making it possible for organizations to adopt modern AI without trading away data privacy or security. The product is aimed at finance, legal, healthcare, and government institutions where sending sensitive data to a third-party model vendor is not an option, and it works by keeping models, data, and infrastructure inside environments the customer already controls, including their cloud tenant, VPC, or on-premise . Engagements are designed against frameworks like HIPAA, SOC 2, GLBA, and ISO 27001, with a typical first production deployment running three to six weeks before expanding to additional workflows. The company is led by founder and CEO Kevin Xie, who studied AI and math at MIT before leaving at 19, published research with Harvard, released a large LLM benchmark, and worked on failure modes in chain-of-thought reasoning . Soren is part of Y Combinator's Fall 2025 batch and is based in San Francisco. The launch lands at a moment when general-purpose AI tools have moved quickly into knowledge work but stalled inside compliance-heavy sectors that cannot expose customer data, case files, or patient records to external training pipelines. Soren's pitch to that audience is concrete operational delivery rather than another model API, with the team handling design, deployment, and ongoing operation so buyers without a dedicated ML org can still put a custom system into production. For founders and operators in regulated verticals evaluating where private AI is finally usable, this is a directly relevant launch to track.