Vertical AI Meets Observability

While modern AI holds great promise, the gap between its promise and practical impact remains substantial. This talk advocates for the importance of vertical AI to help bridge that gap—urging researchers to tailor problem formulations, modeling approaches, data collection, and evaluation methods to concrete downstream tasks. We begin by briefly examining the limitations of existing domain-specific foundation models–for genomics, satellite imaging, and time series–that apply techniques from core AI domains such as vision and NLP with minimal specialization. We then present our ongoing work at Datadog AI Research to develop observability foundation models tailored for forecasting problems associated with production software systems. These efforts highlight the critical role of domain-aware design in moving beyond shiny demos / benchmark leaderboards and toward meaningful AI impact.