Enterprise World Models
Why LLMs need a structured substrate of entities, relationships, and signals to reason about enterprise state.
Co-Founder & Chief Scientist · MIT PhD · Microsoft Research Fellow · Frontier AI
Co-Founder & CEO · ex-Head of AI Products, Microsoft
March 2026
LLMs can draft, summarize, and converse. But the moment you ask them to do something operationally meaningful inside a firm, they hit a wall. The missing layer isn't better prompts or bigger context windows. It's a model of how the enterprise actually works...
Continual Learning Without Catastrophic Forgetting
Methods for adapting foundation models to enterprise domains without losing prior capability.
Co-Founder & Chief Scientist · MIT PhD · Microsoft Research Fellow · Frontier AI
Co-Founder & CEO · ex-Head of AI Products, Microsoft
March 2026
Ninety-five percent of generative AI pilots fail to deliver measurable business impact. The core barrier isn't compute, regulation, or talent. It's that deployed models cannot learn. Enterprise AI stacks — built on static model weights...
From Association to Intervention
Building causal context for language models — moving from pattern retrieval to consequence reasoning.
Co-Founder & Chief Scientist · MIT PhD · Microsoft Research Fellow · Frontier AI
Co-Founder & CEO · ex-Head of AI Products, Microsoft
March 2026
Every modern retrieval-augmented system shares the same quiet assumption: that the best context for answering a question is the text most semantically similar to it. But decisions aren't recall problems. When a central banker asks what happens if we raise the overnight rate...
