I am a Postdoctoral Researcher at Sorbonne University (LIP6, CNRS). My work sits at the intersection of Numerical Linear Algebra, High-Performance Computing (HPC), and Machine Learning.
My core mission is developing efficient algorithms via science and sustainability, focusing on optimizing the trade-offs between accuracy, performance, and energy efficiency for the next generation of supercomputers.
- 🎓 PhD in Applied Mathematics from the University of Manchester.
- 💡 Research Interests: Mixed-precision Algorithms, Tensor Computations, Spatio-Temporal Data Modeling, and Iterative Methods.
Here are some of the key open-source projects I have developed or contributed to, spanning across High-Performance Computing, Time Series Analysis, Clustering, and Numerical Linear Algebra:
| Project | Category | Description |
|---|---|---|
| ⚡ hpc-mix-bench | HPC & Precision | Benchmarking tools for mixed-precision High-Performance Computing (PEQUAN/LIP6). |
| 🛠️ pychop | HPC & Precision | A Python package for simulating low-precision arithmetic and rounding modes in numerical algorithms. |
| 📈 fabba | Time Series | Fast Adaptive Brownian bridge-based symbolic Aggregation for time series discretization and representation. |
| 🤖 llm-abba | Time Series & LLMs | Leveraging Large Language Models (LLMs) with ABBA for advanced time series forecasting and analysis. |
| 🧠 classix | Clustering | A fast, highly scalable, and explainable clustering algorithm based on sorting. |
| 🚀 cusnn & snn | Data Mining | Implementations (including CUDA-accelerated) for Shared Nearest Neighbor (SNN) clustering algorithms. |
| 🔢 blrmat | Linear Algebra | Tools and efficient implementations for Block Low-Rank (BLR) matrix approximations. |
| 🔢 mhodlr | Linear Algebra | Implementations and algorithms for Hierarchical Off-Diagonal Low-Rank (HODLR) matrices. |
- Core Skills: Performance Profiling, Mixed-Precision Computing, Numerical Error Analysis, Parallel Programming.
- Libraries/Frameworks:
CMake,OpenMP,MPI,BLAS/LAPACK,cuBLAS,JAX.
I am actively involved in leading scientific programs shaping the future of computing:
- 🇫🇷 NumPEx (France 2030) - Exa-MA (Oct 2024 - Present) Key project of the European High-Performance Computing Joint Undertaking (EuroHPC). I contribute to developing the software stack and mathematical algorithms for the next generation of Exascale supercomputers.
- 🇪🇺 inEXASCALE (ERC) (Sep 2023 - Oct 2024) An EU-funded project aiming to rethink algorithm design in the Exascale era. The research focuses on the combined effects of multiple sources of inexactness (e.g., lower precision) to develop fast and provably accurate algorithms.
- 🌐 Official Website
- 🐦 Twitter/X
"Mathematics is the music of reason."
"Beneath every cynic, there is an aspiring programmer."