Scholar
Jonas Ngnawé
Google Scholar ID: KwAxSFsAAAAJ
Université Laval, Mila-Quebec AI Institute, Stanford University
Citations & Impact
Citations
13
H-index
3
i10-index
0
Publications
10
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- 1. Two papers accepted at the Reliable ML workshop at NeurIPS 2025: 'Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Adversarial Scheduling' and 'A Guide to Robust Generalization: The Impact of Architecture, Pre-training, and Optimization Strategy'.
- 2. Attending DLRL 2025, the Deep Learning & Reinforcement Learning Summer School in Edmonton.
- 3. Panelist for the 8th Annual Black in AI Workshop at NeurIPS around the theme 'AI Regulation & Fairness in the Generative AI Era'.
- 4. Neptune.ai Neurips 2024 Paper Communication Challenge (Winner).
- 5. Won a best poster award at the '1ère Journée scientifique de l’IID'!
- 6. Our paper on 'Margin Consistency' is accepted at Neurips 2024.
- 7. Appeared in AIMS Alumni of the Week.
- 8. In the Acknowledgments of the book 'Mathematics for Machine Learning' by Prof. Marc Deisenroth, published in 2020.
- 9. AMMI Pioneers video and article in the magazine Jeune Afrique about the launch of the Machine Intelligence Master’s program (AMMI 2018) in Rwanda.
Research Experience
- Currently a Visiting Student Researcher at Stanford University in Fall 2025, at STAIR Lab led by Prof. Sanmi Koyejo. Previously a Google AI resident at the Accra Lab.
Education
- PhD student at Mila-Quebec AI Institute and Université Laval (IID/LSVN lab), supervised by Prof. Christian Gagné (Mila & Université Laval) and co-supervised by Prof. Frédéric Precioso (INRIA & Université Côte d’Azur). Previously a Google AI resident at the Accra Lab, mentored by Yann Dauphin.
Background
- PhD student in Computer Science, with research interests in Trustworthy AI, LLMs/VLMs Safety, Adversarial Robustness, Robust Finetuning, Alignment, Uncertainty Estimation, Robustness to Distribution Shifts, Test-Time Scaling/Adaptation, Active Learning, etc.
Miscellany
- Personal website powered by Jekyll & AcademicPages, a fork of Minimal Mistakes.
Co-authors