CSA STAR Registry

Evara builds a probabilistic programming engine for quantitative reasoning under uncertainty.

Founded by PhD researchers from top universities, published in the leading machine learning and AI venues (NeurIPS, ICML), Evara develops both the science and the software. Our engine builds on our research in probabilistic programming — an emerging branch of AI based on Bayesian inference, orthogonal to LLMs. Our technology encodes domain knowledge and expertise as executable probabilistic models and delivers reliable results even in low-data situations where data-hungry machine learning fails. Learning is embedded in the modeling language itself: a model's assumptions become explicit, tunable parameters that are continuously and automatically corrected as observations and actuals arrive. The result is calibrated estimates and full risk distributions with mathematical guarantees.

The engine integrates natively with Microsoft Excel and transforms it into self-learning software: workbooks update automatically as actuals are written into them. Every number we update is based on rigorous mathematical computation, not black-box generation — inspectable in the workbook, verifiable by construction, incapable of hallucination.

Our product works with any Excel workbook and excels at forecast management, risk quantification, and model validation — for teams accountable for anything from everyday analytics to high-stakes analysis.