PhD at Mila & McGill - Building Time Series Foundation Models One Transformer at the Time
I am a second year PhD student at McGill University and Mila - Quebec AI Institute, supervised by Professors Blake Richards and Irina Rish.
My research focus is on large multimodal models, with my current interests being on how we can adapt them for time series forecasting.
04/2024 - 06/2024, IT Banking Consultant - Forty2 AG
Built retrieval augmented generation pipelines and Avaloq customizations for major European banks.
05/2022 - 08/2022, Technology Analyst - Morgan Stanley
Developed monitoring algorithms to ensure no abuse of database accesses. Algorithms were both statistical and AI based. These ran in the cloud, with Databricks and Snowflake.
04/2022 - 08/2023, System Engineer - Polytechnique Montréal
Responsible for the coordination of a team of 50 students from 11 engineering fields to design a space exploration rover from the ground up to participate in international competitions (URC, CIRC).
03/2021 - 07/2021, Research on classification algorithms to improve cancer detection - Stilla Technologies
Development of 6 dimensional clustering algorithms which would identify rare DNA samples of interest resulting from a PCR while ignoring the rain in different experimental conditions.
Main project: Technical consultant on the Vanuatu domestic submarine cable feasibility study for the Ministry of Foreign Affairs and Trade of New Zealand.
@inproceedings{
roger2026forecasting,
title={Forecasting Emerges from Auto-Regressive Pretraining: Latent Predictive Structure in Language Models},
author={Alexis Roger and Prateek Humane and Zhenghan Tai and Gwen Legate and Andrei Mircea and Vasilii Feofanov and Irina Rish},
booktitle={Forecasting as a New Frontier of Intelligence},
year={2026},
url={https://openreview.net/forum?id=WwxELkT48R}
}
@inproceedings{
roger2026reusable,
title={Reusable Low-Rank Subspaces Explain Why Cross-Modal Transfer Adapts with Tiny Updates},
author={Alexis Roger and Prateek Humane and Zhenghan Tai and Gwen Legate and Andrei Mircea and Vasilii Feofanov and Irina Rish},
booktitle={ICML'26 workshop on CoLoRAI - The 2nd Workshop on Connecting Low-rank Representations in AI},
year={2026},
url={https://openreview.net/forum?id=vu4vYhabO6}
}
@misc{roger2025smallvocabulariesbiggains,
title={Small Vocabularies, Big Gains: Pretraining and Tokenization in Time Series Models},
author={Alexis Roger and Gwen Legate and Kashif Rasul and Yuriy Nevmyvaka and Irina Rish},
year={2025},
eprint={2511.11622},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2511.11622},
}
@inproceedings{
margerie2025image,
title={Image Tiling for High-Resolution Reasoning: Balancing Local Detail with Global Context},
author={Anatole Jacquin de Margerie and Alexis Roger and Irina Rish},
booktitle={AAAI Workshop on Reproducible AI},
year={2025},
url={https://openreview.net/forum?id=nOvSq3KzXU}
}
@inproceedings{riachi2025random,
title={Random Initialization Can{ extquoteright}t Catch Up: The Advantage of Language Model Transfer for Time Series Forecasting},
author={Roland Riachi and Kashif Rasul and Arjun Ashok and Prateek Humane and Alexis Roger and Andrew Robert Williams and Yuriy Nevmyvaka and Irina Rish},
booktitle={1st ICML Workshop on Foundation Models for Structured Data},
year={2025},
url={https://openreview.net/forum?id=UyVsNGlZyn}
}
@misc{roger2025chirpfinegrainedbenchmarkopenended,
title={CHIRP: A Fine-Grained Benchmark for Open-Ended Response Evaluation in Vision-Language Models},
author={Alexis Roger and Prateek Humane and Daniel Z. Kaplan and Kshitij Gupta and Qi Sun and George Adamopoulos and Jonathan Siu Chi Lim and Quentin Anthony and Edwin Fennell and Irina Rish},
year={2025},
eprint={2501.09672},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2501.09672},
}
@inproceedings{
kaplan2024the,
title={The Effect of Data Corruption on Multimodal Long Form Responses},
author={Daniel Z Kaplan and Alexis Roger and Mohamed Osman and Irina Rish},
booktitle={ICML 2024 Workshop on Foundation Models in the Wild},
year={2024},
url={https://openreview.net/forum?id=wP97TI4OEq}
}
Towards Adversarially Robust Vision-Language Models (2024)
Rishika Bhagwatkar, Shravan Nayak, Reza Bayat, Alexis Roger, Daniel Z Kaplan, Pouya Bashivan, Irina Rish Published with presentation in Trustworthy Multi-modal Foundation Models and AI Agents at ICML 2024
@inproceedings{
bhagwatkar2024towards,
title={Towards Adversarially Robust Vision-Language Models: Insights from Design Choices and Prompt Formatting Techniques},
author={Rishika Bhagwatkar and Shravan Nayak and Reza Bayat and Alexis Roger and Daniel Z Kaplan and Pouya Bashivan and Irina Rish},
booktitle={ICML 2024 Next Generation of AI Safety Workshop},
year={2024},
url={https://openreview.net/forum?id=ROHoC5KUxS}
}
@inproceedings{roger2023ethics,
title={Training Large Multimodal Language Models with Ethical Values},
author={Alexis Roger},
booktitle={Master's thesis, Université de Montréal},
year={2023},
url={http://hdl.handle.net/1866/32832}
}
Towards Ethical Multimodal Systems (2023)
Alexis Roger, Esma Aïmeur, Irina Rish Published in AI meets Moral Philosophy and Moral Psychology workshop at NeurIPS 2023.
@inproceedings{
roger2023towards,
title={Towards Ethical Multimodal Systems},
author={Alexis Roger and Esma Aïmeur and Irina Rish},
booktitle={First Workshop on AI meets Moral Philosophy and Moral Psychology},
year={2023},
url={https://neurips.cc/virtual/2023/77050}
}
@inproceedings{10284221,
author={Chaabene, Riadh Ben and Ameyed, Darine and Jaafar, Fehmi and Roger, Alexis and Esma, Aimeur and Cheriet, Mohamed},
booktitle={2023 9th International Conference on Control, Decision and Information Technologies (CoDIT)},
title={A Privacy-Preserving Federated Learning for IoT Intrusion Detection System},
year={2023},
volume={},
number={},
pages={351-356},
keywords={Energy consumption;Privacy;Federated learning;Intrusion detection;Data models;Real-time systems;Internet of Things;Security;Ransomware;Spyware;Federated learning;Security;IoT;Cyberattack},
doi={10.1109/CoDIT58514.2023.10284221}}
@misc{layoun2022aligningmagmafewshotlearning,
title={Aligning MAGMA by Few-Shot Learning and Finetuning},
author={Jean-Charles Layoun and Alexis Roger and Irina Rish},
year={2022},
eprint={2210.14161},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2210.14161},
}
PhD in Artificial Intelligence - McGill University
M.Sc in Computer Science (AI track) - Université de Montréal
B.Sc in Mathematics & Computer Science - École Polytechnique