Sounavo Dey

LLMs trained on text understand language patterns brilliantly. But in domains like manufacturing core challenges are not about language. It's continuous time series data, temperature readings every second, pressure measurements every millisecond, flow rates continuously varying. It's streams of numbers with temporal dependencies, seasonality, and complex autocorrelations.Text models see time series data the way we might try to read a book written in frequencies and waveforms. This session introduces a fundamentally different approach to this problem using time series foundation models and AI agents. Learn how Timeseries agents transforms manufacturing quality control from reactive firefighting to proactive process steering. We'll demonstrate how multivariate forecasting detects process drift before deviations occur, how AI agents reason through root causes without manual investigation, and how manufacturers are reducing defect-to-resolution cycles from weeks to hours.