2048 — C++ Implementation
Recreated the 2048 puzzle with a clean, testable core engine (state, compression/merge, scoring), deterministic seeding, undo, and a lightweight TUI. Unit tests cover chains, no-move states, and spawn logic.
| Apr. 2025 - Sept. 2025 | AI & Quantitative Research Intern at BNP Paribas Global Markets - Data & AI Lab, Frankfurt |
| May. 2024 - Aug. 2024 | Machine Learning Research Intern at ENS Lyon, OCKHAM Team |
| Sept. 2023 - Apr. 2024 | Student Researcher at LIRIS Laboratory, Lyon |
| Sept. 2024 - Mar. 2025 | GRAF Program in Quantitative Finance at ISFA |
| Sept. 2021 - Sept. 2025 | Engineering Student at Centrale Lyon |
| Sept. 2019 - July 2021 | Mathematics and Physics (MPSI-MP*) at Lycée Mohammed VI d'Excellence |
Recreated the 2048 puzzle with a clean, testable core engine (state, compression/merge, scoring), deterministic seeding, undo, and a lightweight TUI. Unit tests cover chains, no-move states, and spawn logic.
Bio-inspired routing with pheromone evaporation/reinforcement balancing exploration vs exploitation; efficient convergence and visualizations across graph topologies.
Contributions improving cross-platform compatibility and reproducibility; added features for performance tracking during the June 2024 coding sprint.
Comparative framework with Merton-type models and Cox processes; Monte Carlo sensitivity to intensity, recovery, and volatility; calibration trade-offs highlighted.
Pricing engine for lookback, Asian, and barrier options under joint stochastic vol & rates; Monte Carlo optimization and calibration to CAC40/EURIBOR data.
Extended Karpathy’s nanochat for multi-turn dialogue; improved preprocessing/tokenization, training stability, and lightweight evaluation/sampling tools.
Integrated a learned discrete latent space (VQ-VAE) with WFC to accelerate constraint propagation and boost diversity while preserving structure.
Entry-level DS project scaffold: data ingestion & EDA, a simple training pipeline, Dockerized service for inference, and CI for lint/tests. Focus on DevOps/MLOps hygiene.
Programming: Python (pandas, NumPy, scikit-learn, PyTorch), SQL, Git, Linux
Quant/DS: Time-series, bandits, optimization (Benchopt), backtesting, risk & pricing
Tooling: Jupyter, Docker, CI/CD (basics)