A selection of projects completed during my studies and beyond.
Modeled a mutual holding situation through mean field games in discrete time. Analyzed how diversification affects the total system — results consistent with the literature: optimal diversification reduces shareholders' risk.
Built algorithmic trading models using Random Forest and XGBoost to predict Forex trends across 9 currency pairs. Feature selection pipeline analyzing 15+ indicators (ATR, ADX, MACD, RSI...). Sharpe ratio of 2.96.
Modeled portfolios of options (calls/puts). Estimated Value-at-Risk using importance sampling, last particle and splitting methods. Auto-regressive Markov chain for conditional law simulation.
Time series classification using Dynamic Time Warping and k-NN on exoplanet WASP-126 b photometric measurements. LSTM and Transformer architectures in PyTorch. 92% improvement in precision (RMSE) with Transformers.
Comparative study of Gaussian Process Regression and XGBoost pipelines for daily mean temperature prediction. Engineered Fourier features to encode seasonality. Best RMSE of 2.65 °C on the test set.
Implemented models to improve ESG under nominal rate shocks and default risk. Hull-White model calibration for inflation (+100% precision). Developed Longstaff-Mithal-Neis model for credit risk on private bonds.
Implemented NSGA II with fast non-dominated sorting. Results corroborated with Benjamin Doerr's article. 100% Pareto set coverage for m=2 and 95% for m=4 on m-LOTZ functions.