Engineering student at Télécom Paris (computer vision, AI and data science), currently on a gap year. Research intern in AI and systematic investing at Université Paris Dauphine-PSL (DRM Finance), working on an event-driven equity strategy that uses local LLMs to turn financial news into trading signals.
Looking for a six-month off-cycle internship in trading or quantitative finance, starting February 2027.
sacha.ryszfeld@dauphine.psl.eu · LinkedIn · Portfolio
| Project | |
|---|---|
| ECB communication and money market rates | Replication of a 2025 Macroeconomic Dynamics paper: hawkish/dovish sentiment indices from ECB press conferences, and HAC regressions showing that the tone predicts EONIA beyond the policy rate. |
| Flip 7 agent | An agent for a push-your-luck card game: expectiminimax search plus a game-level value table learned by approximate policy iteration. Wins 55.8% of 2,000 games against a greedy expected-score strategy. |
| Relational learning on H&M sales | Do graph neural networks on a relational database beat a flat LightGBM? A controlled ablation shows the signal is autoregressive, not relational: the best GNN uses zero hops. Team project on RelBench. |
| White blood cell classification | 13 imbalanced classes from 28,901 microscope images. From handcrafted features to a fine-tuned ConvNeXt-Tiny, which reaches a macro-F1 of 0.726 in 5-fold stratified cross-validation. |
| Image restoration | Denoising and deblurring as variational problems: Tikhonov in closed form and by gradient descent, against Total Variation. |
Python (NumPy, pandas, statsmodels, scikit-learn, PyTorch, PyTorch Geometric, LightGBM, XGBoost, NLTK), SQL, Git, LaTeX.