Pierre Chambet

Problem: Predicting aircraft corrosion in the face of changing weather histories.

Approach: Refusal to over-optimize a model on past data. Creation of a calibrated architecture (LightGBM) for robustness against data drift.

Result: 2nd place overall. The model maintained its performance on unknown test data while competing approaches collapsed. It is this philosophy of robustness that I apply to your operations.

Every airline has a way to absorb disrupted flights. Almost none can say what that reserve really costs them, or what a better one would save.

SmartFleetOptim makes the trade-off visible and testable. Where the loss hides, where you're fragile, how it holds up under stress.

From the airline's fleet structure and costs, it works in two spaces at once: it generates thousands of possible reserve fleets, and it builds a grid of futures those fleets might have to survive. Then it measures how every strategy holds up across every future so you can weigh the trade-offs before you commit.

Operations Research, Optimization, and constrained Simulation.

Data Scientist in the Network Program:

• Worked on predictive modeling and network capacity analysis to support upstream planning decisions.

• Engineered algorithmic models (Python) for operational planning problems.