DCAI : Data-Centric AI
Keywords
Data Quality and Integration, Exploratory Data Analysis, Data Representation, Feature Engineering, Model Training and Optimization, Data Mining, Data Science Pipelines
Heads
Frédéric Flouvat and Sana Sellami
Members
Scientific Objective
The DCAI (Data-Centric Artificial Intelligence) team of the Data Science department of the Computer Science and Systems Laboratory (LIS) focuses on developing innovative methods and tools for data-centric AI.
Our approach aims to maximize the value of available data by optimizing its quality, relevance, and use as input and output for models. With data as our guiding principle, we are also interested in data science pipelines and their automation. We start from the observation that the performance of AI systems relies not only on models but also on how data is collected, prepared, and utilized. In particular, we are interested in how data is used to create models of reality in fields such as medicine, environment, transportation, and road traffic.
We are also interested in using models as tools for representing reality, augmenting data, validating data, and aiding decision-making where analytical solutions are not feasible.