Portrait of Rahma Nouaji

Rahma Nouaji

rahma.nouaji@mila.quebec
PhD - McGill University
Supervisor
Oana Balmau
Research Topics
Distributed Systems
Machine Learning Systems
Website

Publications

A principled approach for energy-efficient training via phase-aware GPU frequency tuning
Miguel Braga
Júlio Pinto
Rahma Nouaji
Olivier Michaud
Bettina Kemme
Oana Balmau
Cláudia Brito
Ricardo Macedo
Modern AI model training imposes unprecedented computational demands, making it a key contributor to datacenter energy consumption. Yet a si… (see more)gnificant fraction of the energy consumed during training does not translate to useful computation due to bottlenecks throughout the training pipeline. We present PAFT, a phase-aware, dynamically adaptable GPU frequency tuning system that reduces energy consumption of training workloads with minimal performance overhead. The key insight behind PAFT is that bottlenecks represent an energy optimization opportunity, rather than purely a performance problem: when GPUs are bound to stall, PAFT opportunistically reduces their clock frequencies to match the pace of bottlenecked devices, saving energy without impacting execution time. PAFT achieves this by continuously monitoring pipeline behavior and applying fine-grained frequency adjustments, adapting to workload and system changes. Experiments conducted on twelve widely used models show that PAFT consistently outperforms all baselines, achieving energy savings of up to 46% with an average overhead of 4%.
2026-09-20
arXiv (preprint)
doi.org
arxiv.org
MinatoLoader: Accelerating Machine Learning Training Through Efficient Data Preprocessing
Rahma Nouaji
Stella Bitchebe
Ricardo Macedo
Oana Balmau
Machine learning (ML) frameworks, such as PyTorch and TensorFlow, rely on data loaders to preprocess data before feeding it to accelerators.… (see more) When preprocessing is inefficiently pipelined, GPUs can remain idle over long periods of time, leading to substantial training delays. For example, PyTorch's default data loaders can cause up to 76% GPU idleness. A key bottleneck is the variability in preprocessing time across samples within the same dataset. Existing data loaders are oblivious to this variability, training all samples uniformly. In this case, a single slow sample can stall the entire batch, causing head-of-line blocking.
2026-04-25
European Conference on Computer Systems (published)
doi.org
arxiv.org