At the crossroads of AI and software engineering research, addressing the challenges of integrating AI/ML components.
As a growing number of applications and even contexts, adopt artificial intelligence (AI), the various AI artifacts that academia and industry produce are under increasing pressure to perform and adapt in changing, large-scale, complex environments. The methods that currently exist for reasoning about software and for guiding software creation and integration are not always adapted to AI applications which is why the new Systemic AI research group is positioned at the crossroads between AI and software engineering research. The team address the challenges of the AI software lifecycle in heterogeneous systems, including (but not limited to) composition, integration, uncertainty and data dependencies of AI/ML components.
One direction we’re exploring relates to the configuration and lifecycle challenges for AI components where we focus on improving the potential long-term reuse and quality of individual algorithmic artifacts. We’re also looking at their relation to data which can be of varying quality in particular in the case of datasets that contain billions of potentially evolving machine-readable facts about the world. This data may be incorrect or incomplete so there’s a strong need for techniques that improve data quality.
We also research how to integrate AI components with heterogeneous components in end-to-end applications, dealing with aspects such as composition and modeling of dynamic behavior. This also concerns deploying, provisioning and testing the resulting applications in containerized environments running in different environments.