Bruno Coussement

Helping organisation improve the way they use their data science platforms through MLOps best practices, self-service infrastructure. This results in decreased time to production, increase uptime and increased feeling of empowerment of teams.

Mostly setting up the AWS landing zone infrastructure, Kubernetes and Neo4J. Secondly, I piloted running containers in trusted execution environments using Scone, and confidential computing nodes on Azure AKS.

Helping organisations create, deploy, maintain data products in the cloud (AWS, Azure and GCP). I'm mostly involved in projects improving machine learning model creation and industrialisation across teams. Data Minded is a Leuven (Belgium) based data engineering consulting firm.

Tensor factorizations for Machine Learning. Representing data using a matrix (similar as you would in an Excel sheet) limits itself to relationships between only 2 objects. Tensors (higher dimensional generalization of matrices/vectors) enable us to model the multiple aspects in a natural way. In the context of recommender systems, modelling Item-User-Location-Time using tensors will yield more valuable insights than limiting yourself to only User-Item interactions. My (broad) goal is to introduce tensor-based algorithms in the machine learning world that naturally leverage their disproportional representational power. Scalable tensor decompositions are vital in achieving this, and thus this is where my focus lies. Other research interests include learned image compression, and speeding op very large feed forward neural nets by adding structure to the weights.

Enabling the France corporate sales team with insights about the French market up to store level. - Analyzing sales data - Reporting about P&G's performance on several levels of granularity (French market up to store level). - Identifying factors that could help grow a particular segment, and P&G market share inside it.

Guiding 7-12 year old children with their first step in tennis.