Hello! I'm Tidiane Camaret Ndir, a machine learning researcher at the University Medical Center of Freiburg, Germany. I'm dedicated to building intelligent systems that have the ability to generalize to unseen settings.

I believe that the future of intelligent systems, especially in medicine, requires models that not only perform complex tasks accurately, but can also take direct human feedback as an input, allowing them to generalize without requiring extensive fine-tuning.

This page is where I explore the methodologies behind building these systems. Thanks for stopping by !

Publications

EEG-CLIP: Learning EEG representations from natural language descriptions

Tidiane Camaret Ndir, Robin T. Schirrmeister, Tonio Ball

Frontiers in Neuroscience, 2025

A multimodal deep learning framework that aligns EEG signals with clinical text reports using contrastive learning. Achieved 85% accuracy in detecting neurological disorders and demonstrated strong zero-shot classification capabilities.

Paper Code

Dynamic Prompt Generation for Interactive 3D Medical Image Segmentation Training

Tidiane Camaret Ndir, Alexander Pfefferle, Robin T. Schirrmeister

arXiv preprint arXiv:2510.03189, 2025

A training strategy combining dynamic volumetric prompt generation with content-aware adaptive cropping for interactive 3D biomedical segmentation. Achieved 2nd place out of 257 teams in CVPR 2025 SegFM3D Competition.

Paper Code

Inferring Behavior-Specific Context Improves Zero-Shot Generalization in Reinforcement Learning

Tidiane Camaret Ndir, André Biedenkapp, Noor Awad

European Workshop on Reinforcement Learning (EWRL), 2024

Proposes jointly learning policy and context representations for zero-shot generalization in reinforcement learning. Demonstrates improved adaptation to unseen environments without additional training.

Paper Code