CV
Download CV (PDF) Download resume (1 page)
Education
- ENS Paris-Saclay - Master MVA (Mathématiques, Vision, Apprentissage), Palaiseau, France (September 2025 - Present)
- France’s top master’s research program in artificial intelligence, computer vision, and machine learning.
- Institut Polytechnique de Paris - Télécom Paris, Engineering Degree (Diplôme d’Ingénieur), Palaiseau, France (September 2023 - December 2026)
- GPA: 3.99/4.0 (top 3% of students). Majoring in applied mathematics, computer science, and artificial intelligence.
- Bachelor of Science, Marrakech and Rabat, Morocco (September 2021 - July 2023)
- Intensive two-year undergraduate program preparing for entrance into France’s top engineering schools.
- Ranked 1st at Lycée Ibn-Timya, Marrakech; awarded a full scholarship to continue studies at Lycée Al-Zahrawi.
Experience
- Harvard AI and Robotics Lab - Visiting Research Intern, Cambridge, MA (April 2026 - Present)
- Investigating connections between continuum mechanics and representation learning, with applications to self-supervised learning and understanding deformation, transport, and optimal control dynamics in flow matching.
- Developing large language model methods for survival and time-to-event prediction on medical data from Harvard hospitals, with a research publication in preparation for Nature Machine Learning.
- Contributing to a research project on geometry-aware in-context learning for multimodal vision-language models, developing spatially informed demonstration retrieval methods to improve spatial reasoning in VQA, and extending the approach to vision-language-action robot control, with a publication in preparation.
Research projects
- Diffusion Policy for Visuomotor Control
- Implemented diffusion-policy architectures for robot action generation conditioned on visual and proprioceptive observations.
- Studied state-of-the-art transformer-based denoising policies for sequential decision-making in robotics.
- Pseudoinverse-Guided Diffusion Models for Inverse Problems
- Implemented a training-free guidance method for diffusion models using pseudoinverse approximations of forward operators.
- Applied the method to image restoration tasks including inpainting, super-resolution, blur, and compression artifacts, and compared against diffusion posterior sampling baselines.
- LLM Agents for Text-Based Game Reasoning (Zork)
- Designed autonomous LLM agents for planning, memory, and interaction in text-based environments such as Zork/Jericho.
- Implemented reasoning loops and action-selection policies to improve exploration in partially observable environments.
- Controlled and Efficient LLM Generation
- Implemented controlled decoding methods for language models, including DFA/FSM-constrained generation for structured outputs and rule-based constraints.
- Studied inference-efficiency techniques for transformers, including KV-cache optimization, LoRA-based adaptation, and FlashAttention-style memory-efficient attention.
Additional projects
- Deep reinforcement learning for video games: DQN and PPO agents for Atari, Super Mario Bros, and MuJoCo simulations.
- Generative architectures reproduced from the literature: WaveNet (raw audio), SinGAN (single-image GANs), and Listen, Attend and Spell (speech recognition).
- Structured graph learning via spectral constraints: differentiable graph recovery from time series.
- Quantifying extrinsic curvature in neural manifolds with differential-geometry tools and variational autoencoders.
- Interpretable hyperspectral satellite image unmixing combining deep learning and plug-and-play optimization.
- Kaggle molecular captioning challenge: fine-tuned LLaMA, Gemma, and RoBERTa with graph neural networks for molecular description generation.
Extracurricular experience
- Treasurer, Télécom Paris Robotics association, Palaiseau, France (January 2024 - August 2025)
- Managed the association’s annual budget, sponsorships, and financial compliance for robotics projects and events.
- Two-time participant in La Coupe de France de Robotique, contributing object-recognition, path-planning, and reinforcement-learning modules in C++, Python, and ROS 2.
Skills
- Programming languages: C, C++, Python, Java, JavaScript/TypeScript, HTML, CSS, React.js
- Frameworks, AI libraries and tools: PyTorch, TensorFlow, NumPy, scikit-learn, OpenCV, Git, Linux, Bash/Shell
- Robotics frameworks: ROS 2, Pinocchio, MuJoCo, Arduino, Gymnasium
Honors and awards
- Ranked 1st in class at Lycée Ibn Timiya, Marrakech.
- Full merit scholarship, Lycée Al-Zahrawi, Rabat.
- Government scholarship (France-Morocco), granted jointly by the French and Moroccan governments for higher studies in Paris.
- Semi-finalist, French Debating Association of the Grandes Écoles of Paris, representing Télécom Paris.
