TENSLORA: Tensor Alternatives for Low-Rank Adaptation
Marmoret, et al. (2025). "TensLoRA: Tensor Alternatives for Low-Rank Adaptation." In 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE.
CV
Jonathan Lys
Summary
Second year PhD student at IMT Atlantique. Previously research intern in the Criteo AI Lab. Computer Vision, Natural Language Processing and Deep Learning enthusiast. I like to optimize the training and the generation of large language models and diffusion models.
Work Experience
- Visiting Scientist2024 - 2025Sony Europe, Stuttgart, GermanyTwo visits to EUREC, total of 3 months.
- Worked on conditional generation for text diffusion models.
- Studied looped AR LLMs for reasoning tasks.
- Research InternApril 2024 - October 2024Criteo AI Lab, Paris, FranceStudied diffusion models, for image background generation.
Supervised by Jaspreet Sandhu, Thibaut Issenhuth, Ludovic Dos Santos- Trained large-scale diffusion models up to 6B parameters.
- Developed an ad background generation pipeline, reducing object extension by 10×, achieving lowest FID and fast inference.
- Worked on billions of product images, leveraging PySpark and distributed training.
- Research InternSummer 2023IMT Atlantique, BRAIN Team, Brest, FranceStudied parameter-efficient fine-tuning methods for SAM and attention-based data augmentation techniques. Supervised by Bastien Pasdeloup.
- Matched the performance of full fine-tuning with a custom LoRA method.
- Improved the performance of CNNs through attention-based data augmentation.
- Developer InternFebruary 2022French Polynesia’s government IT Department, Papeete, French PolynesiaDevelopment of a Web application for network management, automation of database migrations
Education
- Ph.D Student in Computer Science2024 - 2027IMT Atlantique, BRAIN TeamFocusing on discrete diffusion models for NLP applications and EEG models.
Supervised by Vincent Gripon, Bastien Pasdeloup, Axel Marmoret. - Engineering Degree2021 - 2024IMT AtlantiqueMathematical and Computational Engineering specialization.Topics: Machine Learning, Deep Learning, Computer Vision, Signal Processing, Optimization
- M.Sc. Engineering Sciences and Applications2023 - 2024University of Rennes 1Double degree program with Rennes 1 and CentraleSupélec.
Signal and Image processing major.
Skills
Programming Languages
- Python
- Julia, R
- Java, SQL
- C++
- Matlab
Machine Learning
- numpy
- pandas
- sklearn
- Statistical and probabilistic models
- Data analysis
Deep Learning
- PyTorch (distributed)
- TensorFlow/Keras
- CNNs and Transformers training
- triton
Hugging Face libraries
- transformers
- diffusers
- datasets
- accelerate
Distributed Systems
- Multi-gpu/multi-node training
- SLURM
- Hadoop
- Spark
Soft Skills
- Teamwork
- Communication
- Agile methods
- Problem solving
Publications
REVE: A Foundation Model for EEG – Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects
El Ouahidi, et al. (2025). "REVE: A Foundation Model for EEG Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects." In Advances in Neural Information Processing Systems (NeurIPS 2025).
FICUS: Few-shot Image Classification with Unsupervised Segmentation
Lys, et al. (2024). "FICUS: Few-shot Image Classification with Unsupervised Segmentation." In 2024 32nd European Signal Processing Conference (EUSIPCO)(pp. 1791-1795). IEEE.
Languages
French (Native), English (C1, IELTS 7.5), Mandarin (B2.1, HSK3 equivalent), notions of Spanish and German
Interests
- Music (violin for 12 years)
- Competitive chess (World top 0.3% online)
- Photography
Associations
Secretary, IMTA PhD Student Association; President, IMTA Photography Association (until 2023); President, IMTA Short Film Association (until 2023); Officer, IMTA Sound & Light Association (until 2023); Member, IMTA Arts Association (until 2023); Member, IMTA Chess Association