Jérémy Chaze

Developed and deployed a predictive ML model for morning fixings using LSTM networks, outperforming Random Forest baselines by capturing multi-period temporal dependencies. Designed and backtested statistical arbitrage strategies on European equity indices; automated the full pipeline from data collection to signal generation. Documented model architecture, assumptions and limitations to facilitate handover and reuse by the team.

Built Physics-Informed Neural Networks (PINNs) to simulate energy production from turbulent fluid flows, combining deep learning with physical constraints. Ran high-dimensional optimization routines to maximize turbine efficiency; validated models against experimental data.