Research Blog

What can time series researchers learn from the success of large language models, and what makes sequential numerical data fundamentally different from text? We discuss our experiences building TSFounder.

How our ThermoShield project uses generative models and physics-informed optimization to identify novel composite materials that passively reduce greenhouse heat load by up to 18°C.

We introduce RAG-Impute, a retrieval-augmented approach to missing value imputation in multivariate time series that leverages a large corpus of historical patterns to guide reconstruction.