Research Scientist Intern - Post-Training (RLHF)

This six-month internship focuses on designing and applying RLHF methods to diffusion-based text-to-image models. You will conduct literature reviews, implement RLHF/DPO techniques, and participate in large-scale model fine-tuning. Design rigorous ablations, establish evaluation metrics, and analyze model performance. Document findings, prepare technical reports, and contribute to an ambitious open-source project and community. Collaborate with researchers across ML, CV, and NLP; opportunities to publish results. Paris-based with hybrid flexibility; on-site work required as necessary; strong background in Python and DL frameworks essential, with preference for PhD or MSc candidates and prior RLHF experience.