Liquid AI

Liquid AI is a spin-off from MIT CSAIL that builds Liquid Foundation Models, an approach to foundation models designed from first principles rather than on the standard transformer architecture. The models are built to run efficiently across a range of deployment targets, from data centres to edge devices, and are released as white-box models. The company's technical work spans foundation models, model efficiency, adaptive systems, non-transformer architectures and on-device inference.

The team combines engineering, research and business leadership, and describes itself as a group of engineers, researchers and business leaders. Members are treated as domain experts with high trust to execute autonomously, and are expected to work on the critical path, call out problems early and step outside their immediate remit where needed.

Culture signals emphasise autonomy, purposeful ownership and a merit-based evaluation of ideas, with results valued above hierarchy. The company states that it operates without playbooks, preferring to rebuild from scientific fundamentals, and that transparency is central to collaboration - describing white-box models within a white-box organisation. Stated norms also include prioritising colleagues' wellbeing and aligning company goals with individual professional growth.