Engram develops an AI memory architecture that separates a model’s reasoning/inference layer from its memory layer to enable online continual learning and real-time updates without full retraining. The company says this approach allows models to adapt to user preferences, conversation history and newly arriving data within seconds or hours and to dramatically reduce token-related operating costs. Engram reports strategic partnerships with Microsoft and Notion and claims some implementations can cut AI-related costs by a factor of 10 to 100. It was founded last October out of Stanford’s AI lab by Dr. Dan Biderman, includes Dr. Natalie Biderman among its early team, and counts advisers such as Andrej Karpathy and Pieter Abbeel. Engram remains a lean team of 13 employees and states it is already generating significant revenue through contracts with large customers. The company raised $98 million at a $600 million valuation in the reported financing.

Total raised

$98M

Funding rounds

1

Latest round

Equity

Latest activity

Jun 2026