<aside> ⛰️
Time series are the operating language of the physical economy, yet today's LLMs cannot understand them. With OpenTSLM (ICML '26), we introduced scalable Time-Series Language Models: models that perceive signals, connect them with text, and explain their decisions. We are now building the operational intelligence layer from Europe, powering personalized medicine, autonomous enterprise, factories, and energy systems.
</aside>
<aside> <img src="/icons/cloud_gray.svg" alt="/icons/cloud_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/table_gray.svg" alt="/icons/table_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/chemistry_gray.svg" alt="/icons/chemistry_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/playback-fast-forward_gray.svg" alt="/icons/playback-fast-forward_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/reference_gray.svg" alt="/icons/reference_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/groups_gray.svg" alt="/icons/groups_gray.svg" width="40px" />
</aside>
The US has software scale. China has a manufacturing scale. Now Europe needs the AI multiplier for its industries.
Europe's industries generate time series from factories, hospitals, grids, and machines. These signals already drive operational decisions, but today they are fragmented and underused. Realizing their full potential is Europe's competitive edge.
Today's LLMs are capable across text, images, and audio, yet they remain unable to reason over temporal signals. Time-series foundation models perceive signals but cannot use text context. Two families of models, each missing half the problem. OpenTSLM bridges this gap: one model that perceives raw time-series and reasons over them in natural language.