2026Under reviewarXiv preprint
Stable FP4 Training via Transposition-Invariant Block Quantization
Mehdi Rahimifar, Amin Darabi, Mehran Taghian Jazi, Xing Huang, Yao Wang, Zhijun Tu, Yufei Cui, Yunke Peng, Hongliang Li
ML Researcher & Engineer
Amin Darabi
ML researcher and engineer specializing in efficient large-scale training, GPU kernel development, and low-precision training.
Montreal, Canada
Research interests
- Low-precision training (FP8 / FP4)
- Efficient large-scale training
- GPU kernels (Triton / CUDA)
- Distributed multi-node training
- Foundation models for time series & neuroimaging
- Test-time adaptation
Currently
Dec 2024 – Present
Efficient AI Researcher (Intern)
Huawei Canada
Low-bit training (FP8 / FP4), model pruning, and test-time adaptation. GPU kernel development in Triton and CUDA. Pre-training large transformers in multi-node, multi-GPU distributed setups.
Sep 2023 – Present
Graduate Research Assistant
Mila – Quebec AI Institute
Research on time-series and neuroimaging foundation models at the CERC-AAI Lab, supervised by Professor Irina Rish.
Selected publications
2024NeurIPS Workshop on Time Series in the Age of Large Models
General-Purpose Brain Foundation Models for Time-Series Neuroimaging Data
Mohammad Javad Darvishi Bayazi, Hena Ghonia, Roland Riachi, Bruno Aristimunha, Arian Khorasani, Md Rifat Arefin, Amin Darabi, Guillaume Dumas, Irina Rish
2024BMC Bioinformatics 25, 241
AFITbin: a metagenomic contig binning method using aggregate l-mer frequency based on initial and terminal nucleotides
Amin Darabi, Sayeh Sobhani, Rosa Aghdam, Changiz Eslahchi