Jama Mohamud

Mila - Montreal Institute for Learning Algorithms

6666 rue Saint-Urbain

Montréal, QC H2S 3H1, Canada

hussein-mohamu.jama 🌀 mila.quebec

Bio

Up until 2024, I was a nomadic AI research engineer working at different institutions and companies, having traveled to over 20 countries, I have now settled in Montreal working at Mila Quebec AI institute. Before Mila, I was research Fellow at German Research Center for Artificial Intelligence (DFKI). Prior to DFKI, I was a Google DeepMind scholar working with the remarkable Ulrich Paquet and Willie Brink on understanding why deep learning works, particularly studying why pretraining models are so effective. And before this, I was working on privacy preserving self-supervised representation learning with Moustapha Cisse.

💡 I have also recently started PhD at the University of Montreal, supervised by Mirco Ranavalli and Yoshua Bengio. My research primarily revolves around the intersection of AI and neuroscience. I am interested to understand how the human brain functions by developing general-purpose models for the brain. I particularly work on deep representation learning, graph neural networks and language models and their applications to human brain, drug discovery, and healthcare.

Additionally, I’m exploring foundational models, with a particular emphasis on fine-tuning large language models for downstream tasks like reasoning and code generation.

Community work

Beyond my research, in my free time, I’m passionate about contributing to open-source projects and machine learning for societal good. Particularly, I would like to study the effects of weather like rising temperatures, floods, etc., on various geographical scales. If you have an exciting project or are seeking collaboration or have a geo-spatial problem, please don’t hesitate to reach out. I’m always eager to learn and help. In the past, I have:

Served as a TA for two consecutive years at Neuromatch Academy for the Deep learning Summer School
Served as a TA at African Institute for Mathemetical Science to help next cohort AMMI students on foundational machine learning courses.
Contributed to several open-source projects, including:
- Gflownets Library a machine learning framework for probabilistic and generative modelling
- Graphium Library for scaling molecular GNNs to infinity
- Jraph - A library for graph neural networks in jax.
Supported the Deep Learning Indaba (DLI), the largest AI conference in Africa, by contributing to the practicals for three consecutive years:

selected publications

Towards Foundational Models for Molecular Learning on Large-Scale Multi-Task Datasets

Dominique Beaini, Shenyang Huang, Joao Alex Cunha, and 32 more authors

International Conference on Learning Representations (ICLR), 2023

Bib PDF

@article{Mohamud2023FoundationalModels,
  title = {Towards Foundational Models for Molecular Learning on Large-Scale Multi-Task Datasets},
  author = {Beaini, Dominique and Huang, Shenyang and Cunha, Joao Alex and Li, Zhiyi and Moisescu-Pareja, Gabriela and Dymov, Oleksandr and Maddrell-Mander, Samuel and McLean, Callum and Wenkel, Frederik and Müller, Luis and Mohamud, Jama Hussein and Parviz, Ali and Craig, Michael and Koziarski, Michał and Lu, Jiarui and Zhu, Zhaocheng and Gabellini, Cristian and Klaser, Kerstin and Dean, Josef and Wognum, Cas and Sypetkowski, Maciej and Rabusseau, Guillaume and Rabbany, Reihaneh and Tang, Jian and Morris, Christopher and Koutis, Ioannis and Ravanelli, Mirco and Wolf, Guy and Tossou, Prudencio and Mary, Hadrien and Bois, Therence and Fitzgibbon, Andrew and Banaszewski, Błażej and Martin, Chad and Masters, Dominic},
  year = {2023},
  journal = {International Conference on Learning Representations (ICLR)},
}

Exploring the Application of Perceivers to Graph Representation Learning

Jama Hussein Mohamud, Thomas Makkink, Arnu Pretorius, and 1 more author

In , 2022

Bib

@inproceedings{Mohamud2022GraphPerceivers,
  title = {Exploring the Application of Perceivers to Graph Representation Learning},
  author = {Mohamud, Jama Hussein and Makkink, Thomas and Pretorius, Arnu and Fu, Yunguan},
  year = {2022},
  journal = {Research done at InstaDeep - Ongoing},
}

Fast Development of ASR in African Languages using Self-Supervised Speech Representation Learning

Jama Hussein Mohamud, Lloyd Acquaye Thompson, Aissatou Ndoye, and 1 more author

European Chapter of the Association for Computational Linguistics (EACL), 2021

Bib PDF

@article{Mohamud2021ASR,
  title = {Fast Development of ASR in African Languages using Self-Supervised Speech Representation Learning},
  author = {Mohamud, Jama Hussein and Thompson, Lloyd Acquaye and Ndoye, Aissatou and Besacier, Laurent},
  year = {2021},
  journal = {European Chapter of the Association for Computational Linguistics (EACL)},
}

Self-supervised Private Representation Learning

Jama Hussein Mohamud, Mitiku Yohannes, and Moustapha Cisse

Thesis - available upon request, 2020

Bib

@article{Mohamud2020PrivateLearning,
  title = {Self-supervised Private Representation Learning},
  author = {Mohamud, Jama Hussein and Yohannes, Mitiku and Cisse, Moustapha},
  year = {2020},
  journal = {Thesis - available upon request},
}

Poverty Level Characterization via Feature Selection and Machine Learning

Jama Hussein Mohamud, and Omer Nezih Gerek

IEEE Symposium on Signal Processing and Information Technology, 2019

Bib

@article{Mohamud2019PovertyLevel,
  title = {Poverty Level Characterization via Feature Selection and Machine Learning},
  author = {Mohamud, Jama Hussein and Gerek, Omer Nezih},
  year = {2019},
  journal = {IEEE Symposium on Signal Processing and Information Technology},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=b-6wf6wAAAAJ&citation_for_view=b-6wf6wAAAAJ:u5HHmVD_uO8C},
}