Jules Decaestecker
Ai Research Intern @Thales
Signup · Get unlimited contacts
WORK HISTORY
Ai Research Intern @Thales
Paris, FR
Implementation of Mamba models for high-resolution semantic image segmentation.Designed a custom Mamba architecture (PyTorch).Conducted a comparative study with Transformers (ViT, Swin), focusing on accuracy and computational efficiency.Optimized the architecture for inference on resource-constrained embedded systems (NVIDIA Orin AGX), using TensorRT acceleration.Ported and benchmarked the model on embedded hardware, evaluating performance/latency trade-offs.Carried out a literature review of State Space Models applied to vision, and assessed their industrial relevance.
EDUCATION
IMT Atlantique
Baccalauréat
Lycée Saint-Vincent de Senlis
Baccalauréat scientifique
Université de Tokyo
Transfert de crédits
ENS Paris Saclay
Master MVA
ABOUT JULES DECAESTECKER
As an AI Engineer and a graduate of the prestigious MVA Master\'s program (ENS Paris-Saclay), I am passionate about designing and optimizing cutting-edge artificial intelligence models. Currently an AI Research Intern at Thales, I thrive on solving complex, real-world challenges. My expertise is built on three key pillars: Computer Vision: Developing innovative models (Mamba, Transformers) and optimizing them for high-performance inference on embedded systems (NVIDIA Jetson, TensorRT). Generative AI & LLMs: Fine-tuning Large Language Models (LoRA) and deploying production-ready RAG pipelines for practical applications. Reinforcement Learning: Applying advanced algorithms (PPO, DQN) to solve complex problems in robotics (bipedal robot stabilization) and healthcare (HIV treatment optimization). I am actively seeking a full-time position as a Machine Learning Engineer, AI Scientist, or R&D Engineer, available from October 2025. I am eager to contribute my skills to ambitious and impactful projects. Please feel free to reach out!
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.