Jean-Luc Mahoromeza

Vinci Airports needed a reliable system to forecast passenger traffic over a 12-month horizon across multiple operational levels, from flights to terminals and entire airports.

I was responsible for designing and implementing a multivariate hierarchical forecasting engine capable of ensuring consistency between granular and aggregated projections.

I developed a production-ready framework combining state-space models, Bayesian hierarchical modeling, and online learning techniques. The system generates long-term forecasts while updating daily as new data becomes available, allowing projections to dynamically adjust to recent demand signals and structural changes.

The solution strengthened long-term demand visibility while maintaining short-term responsiveness. It enabled more reliable capacity planning, reduced manual forecast revisions, and provided quantified uncertainty intervals to support operational and strategic decision-making.

Led the design and delivery of data-driven solutions within the Cloud Enterprise Solutions team, supporting clients in translating business challenges into scalable machine learning systems on Google Cloud Platform (GCP). I was responsible for the end-to-end lifecycle of ML projects: problem framing, statistical modeling, experimentation, validation, and production deployment. Solutions were built using GCP services, ensuring scalability, security, and operational robustness.

I advised clients on ML architecture design, implementing continuous training and deployment pipelines to enable model lifecycle management in production environments. This included infrastructure setup, monitoring, and governance to ensure reliability over time.

As Team Lead, I contributed to structuring data science initiatives, mentoring team members, and supporting pre-sales activities by defining technical solutions aligned with client strategy and cloud constraints.

I design high-reliability forecasting systems for organizations operating under structural uncertainty and complex demand dynamics.

At CNRS and LPSM, I work on advanced multivariate and hierarchical time-series models capable of delivering coherent forecasts across multiple aggregation levels. My expertise lies in probabilistic forecasting and uncertainty-aware modeling, ensuring that predictions remain stable, interpretable, and decision-ready.

I develop adaptive state-space and Bayesian frameworks that update continuously as new data becomes available, allowing long-term projections to remain responsive to short-term signals. This combination of multi-horizon forecasting and online learning is particularly suited for large-scale operational environments.

My positioning is deliberately niche: I focus on robust forecasting architectures for high-stakes systems where uncertainty must be quantified and controlled, not ignored.

I bridge mathematical rigor and production-grade implementation, transforming theoretical modeling into scalable forecasting engines deployed on real operational data.