Numerical Computing Researcher / Expert – AI Infrastructure

Numerical Computing Researcher / Expert – AI Infrastructure
Location: Paris, France
Team: AI Infrastructure R&D Lab
Position Overview
Our client in Paris is seeking a highly skilled Numerical Computing Specialist to join their cutting-edge AI Infrastructure team. You will be part of a world-class R&D group dedicated to developing next-generation infrastructure for large-scale AI models. In this role, you will address core numerical computing challenges to enable efficient, accurate, and scalable model training and inference.
They welcome candidates at multiple levels, including exceptional recent PhD graduates and seasoned technical experts.
Key Responsibilities

• Conduct advanced research in numerical computing for large model training and inference
• Design and implement innovative techniques, such as:
  • Sparse matrix computation
  • Low-rank approximation and tensor decomposition
  • Mixed precision computation and stability analysis
  • Quantization and dimensionality reduction
• Collaborate closely with system, compiler, and parallel computing experts to integrate innovations into real-world systems
• Publish in top-tier conferences or contribute to impactful industrial solutions

Qualifications
Required:

• Strong background in numerical linear algebra, scientific computing, or applied mathematics
• Proven ability to implement and analyze algorithms involving matrix/tensor computation
• Strong coding skills in Python and/or C++, with experience using AI frameworks such as PyTorch, TensorFlow, or JAX
• Ability to read and write academic papers; fluent English

Preferred:

• PhD in computer science, applied mathematics, or related field (or equivalent industry experience)
• Research experience in one or more of the following:
  • Sparse/dense matrix algorithms
  • Mixed precision or adaptive precision methods
  • Model compression or efficient representation
• Experience with AI compilers, numerical kernels (e.g., cuBLAS, CUTLASS), or HPC libraries
• Publications in leading conferences such as NeurIPS, ICML, ICLR, SC, SIAM, etc.