Bilal FAYE
PhD in Artificial Intelligence | Researcher | Educator | AI Vision & NLP Specialist.
📍 Saint Denis, France | ✉️ biljolefa@gmail.com | LinkedIn
PhD in Artificial Intelligence | Researcher | Educator | AI Vision & NLP Specialist.
📍 Saint Denis, France | ✉️ biljolefa@gmail.com | LinkedIn
Institution: GREYC Lab, UMR CNRS 6072 – Caen
Title: Toward Frugal Multimodal Models: Leveraging Prior Knowledge for Efficient Learning
Abstract: Incorporating prior knowledge can significantly reduce the need for large-scale training data and heavy parameterization, while preserving high performance. Within a multimodal framework, this principle enables the design of lightweight and efficient models. Two key examples illustrate this idea: OneEncoder, a frugal multimodal encoder based on contrastive learning that aligns diverse modalities (text, image, audio, video) without relying on massive paired datasets; and LightMDETR, a streamlined adaptation of the MDETR model for open-vocabulary object detection, which enables efficient generalization to unseen categories with reduced computational demands.
Event: Paris Generative AI Autumn School
Collaboration: Yunhao Tang (Meta FAIR)
Focus: Offline alignment, reinforcement learning for AI model optimization without online interaction, supervised fine-tuning (SFT), preference learning, and safe alignment.
Topic: Addressing hallucinations and domain bias in LLMs for medical queries, using sepsis as a case study. Discussed methods to improve reliability and robustness in healthcare applications.
Topic: Role of attention mechanisms in multimodal fusion and graph processing. Covered their theoretical basis and practical applications in integrating complex data.
Audience: Master 2 DataScale students
Content: Practical session on LLM training and fine-tuning using Hugging Face Transformers. Focused on optimization strategies and deployment best practices.
Topic: Explored normalization strategies in deep learning, focusing on methods that use prior knowledge for better adaptability and efficiency.
Audience: General public and students
Content: Fundamentals of Artificial Intelligence, applications across domains, and a look at emerging trends in the field.