Klarity Workbench (WB)

The Workbench refers to a classical engineering workbench for AI development : mainly an MLops toolchain. Inside WB we use tools that generated output metrics use to build artefact in KD.

2 kind of artefacts :

Artefact build manually exemple ( UC description, ODD requierement, ... ) Artefact build from outputs of tools ( explainability, ... )

Artefact build manually exemple

A configuration of the artefact A sample of payload build manually

Artefact build from outputs of tools

Which tools will be use to generate metrics on what "scope"

apply DQM on dataset X with a Configuration Y that generate a metrics Z

How we build the artefact payload

Artefact A use metrics Z1, Z2, Zn in that way exemple Metric Z1 = name, luminosity of each images in the dataset Metric Z2 = explainability of Model M1 Metric Z3 = anomaly probability by M1 Artefact is a list of datasample with column luminosity and anomaly probability, with image and explaimability overlay

So we need a list of metrics which define

What is the scope What is the configurability of the metrics Which tools generate the metrics What is the structure of the metrics.