AI and quantitative finance
Engineering student at Télécom Paris, on a gap year as a research intern in AI and systematic investing at Université Paris Dauphine-PSL.
Looking for a six-month off-cycle internship in trading or quantitative finance, starting February 2027.
GitHubLinkedInsacha.ryszfeld@dauphine.psl.eu
I am an engineering student at Télécom Paris, where I studied probability, statistics, optimization and machine learning, from linear models to deep generative models. Before that, I prepared the engineering school entrance exams in preparatory classes (PSI*) at Lycée Claude Bernard.
I am spending my gap year at Université Paris Dauphine-PSL, as a research intern in AI and systematic investing with Prof. Serge Darolles. I work on a trading strategy for European equities that reads financial news with large language models, and on the execution side that makes it run every day: order sizing, rebalancing and monitoring.
I am also starting to contribute to Prof. Darolles’ research on dynamic portfolio allocation, with score-driven models that update the portfolio weights directly instead of first estimating expected returns and covariances, and their extension to large portfolios.
Alongside the internship, I follow the Empirical Finance course of Dauphine’s Master 222 (portfolios and factor models), and I study derivatives and fixed income on my own.
Outside of work, I enjoy playing the guitar, swimming, and automating bits of my daily life with AI.
A replication of Czudaj and Nguyen (2025). I score the tone of every ECB press conference with a hawkish/dovish dictionary, then test whether that tone predicts the overnight interbank rate beyond the policy rate itself.
1.15pass-through of the policy rate to EONIA, against about 1.14 in the paper
An agent for the two-player card game Flip 7, where each turn you either draw another card or bank your points. It combines an expectiminimax search over the current round with a table of game-winning probabilities learned by approximate policy iteration.
55.8%of 2,000 games won against a greedy expected-score strategy
Do graph neural networks on a relational database beat a flat table with LightGBM? On H&M weekly article sales, a controlled ablation shows that the signal is autoregressive, not relational, and that the best graph model uses no message passing at all.
0 hopsis the best graph model, and LightGBM beats every graph configuration
Classifying microscope images of white blood cells into 13 very imbalanced classes, from a baseline on handcrafted features to a fine-tuned ConvNeXt-Tiny with XGBoost on its embeddings, validated by 5-fold stratified cross-validation.
0.726macro-F1 of the fine-tuned ConvNeXt-Tiny in 5-fold cross-validation
Denoising and deblurring formulated as optimization problems, solved in closed form in the Fourier domain and by gradient descent in PyTorch, comparing a quadratic smoothness prior with Total Variation.
+1.5 dBof PSNR for Total Variation over Tikhonov in deblurring, and SSIM 0.67 against 0.58
Aug 2026 to present
Experience
Université Paris Dauphine-PSL, DRM Finance
Research internship supervised by Prof. Serge Darolles, on trading strategies that use large language models to read financial news.
Sep 2025 to present
Experience
Lycée Claude Bernard, Paris
Weekly oral examinations of preparatory class students: assessing their technical reasoning and giving immediate feedback.
2024 to 2028
Education
Télécom Paris, Institut Polytechnique de Paris
First year of the master’s cycle completed, gap year in 2026-2027. Coursework in probability, statistics, linear models, optimization for machine learning, deep learning, generative models (GANs, VAEs, diffusion) and computer vision.
2021 to 2024
Education
Lycée Claude Bernard, Paris
Intensive mathematics and physics for the competitive entrance exams to French engineering schools. Ranked in the top 10% of the Mines-Ponts exam (365th of 3,664).