Master's student at Sorbonne University - Paris.
Interested in Machine Learning, Deep Learning, Computer Vision, and Natural Language Processing, with a particular focus on applying AI to real-world problems.
Developed a medical imaging pipeline for brain tumor classification and segmentation, with a web interface deployed on AWS EC2.
- Methods: ResNet18, U-Net, MONAI, Transfer Learning
- Results: 0.91 Macro F1 (classification) · 0.72 Mean Dice (segmentation)
- Tools: PyTorch, Docker, AWS EC2
Built a multi-label audio classification pipeline for 234 bird species using 35K+ recordings and 490 engineered acoustic features.
- Methods: MFCCs, Spectral Features, Logistic Regression, SVC, XGBoost
- Results: 0.92 ROC AUC
Combined collaborative filtering, game clustering, and LLM-generated explanations to explore interpretable personalized recommendations.
- Methods: k-NN, NNMF, K-means, LLMs
- Evaluation: ROUGE, BLEU
Benchmarked lexical and neural retrieval methods on 298K Stack Overflow Q&A pairs.
- Methods: BM25, TF-IDF, IBM Model 1, HNSW, Sentence Transformers
- Results: Recall@100 improved from 0.20 to 0.57
Explored traditional NLP and deep learning approaches for speaker identification and movie-review sentiment classification.
- Methods: Topic Models, RNNs, BERT, RoBERTa, Fine-tuning
- Results: F1 improved from 0.62 to 0.81