Helping AI models to meet the real world

MIT researcher Devavrat Shah has developed a new approach to AI that tackles an unglamorous but crucial problem: helping companies make better decisions with their own operational data. Rather than training on text and images like most AI models, Shah's system works with structured data—the kind of spreadsheets and databases that companies actually rely on. His method draws on graphical models, like those used by GPS systems to convert sparse satellite data into accurate position fixes, but applies them to forecasting demand, optimizing supply chains, and planning business operations. Shah co-founded Ikigai Labs in 2019 to commercialize this research. The company was recently acquired by Celonis, a firm specializing in digitizing operations for over one thousand large companies worldwide. Shah is now chief scientist at Celonis, positioning the company to deploy Ikigai's models directly atop their digitized operational platforms, enabling enterprises to simulate options, predict optimal strategies, and forecast decisions at unprecedented scale.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton