When Models Query Models

Explore a conference talk from EuroPython 2022 that delves into the design of large-scale engineering systems using various numerical models. Learn about a uniform modeling interface that encapsulates these models in notebooks, controlling their creation, execution, and result querying. Discover how numerical solvers are embedded into Docker containers, providing isolated and reproducible environments with language-agnostic REST APIs. Understand the concept of a model registry for efficient queries and how the overall system is represented as a collection of interacting models. Gain insights into design optimization through the execution of model dependency trees, studying parameter changes and their optimization. Examine the model query mechanism allowing notebook models to interact, represented by a graph with suitable processing algorithms. Learn about model resolution order derivation and caching for efficient execution. Explore the open-source technologies used in this modeling framework, including pydantic, Fast API, Jupyter, papermill, scrapbook, Docker, Openshift, MongoDB, and Redis. Gain valuable knowledge on good practices and design decisions encountered during the development process of this innovative approach to large-scale engineering system design.

When Models Query Models - presented by Michal Maciejewski