AI Benchmark Specialist

Benchmark ML Engineer

Bull is a company that develops high-performance computing and AI solutions, designing powerful, sustainable, sovereign digital solutions for states, industries and the scientific community.

You will work within the Applications and Performance team (over 30 engineers) whose main mission is to respond to HPC & AI tenders, planning and guaranteeing the performance of customers' scientific applications on proposed supercomputers.

Job Responsibilities

• Prepare, execute, and analyze AI benchmark applications, primarily training or inference benchmarks from the MLPerf suite (Training and Inference: Datacenter).
• Benchmark Analysis:
  • Conduct literature review on the target application.
  • Explore source code when available.
  • Match hardware architecture with hyper-parameters.
• Benchmark Preparation:
  • Set up software environment in containers.
  • Write training and job scripts.
  • Prepare datasets and perform hyperparameter search.
  • Document procedures and results.
• Benchmark Execution and Performance Estimation:
  • Run tests on multi-node, multi-accelerator clusters.
  • Analyze latency, throughput, accuracy vs performance, and scaling efficiency.
  • Report findings and generate estimates for non-existent future systems.

Qualifications

• Relevant degree in higher education or university.
• Several years of experience in AI.
• Strong sense of curiosity and initiative; ability to work independently and collaborate in a team.
• Solid understanding of deep learning fundamentals and experience with classic neural network architectures (MLP, CNN, RNN), Transformers, and LLMs.
• Experience with Docker or Singularity containers.
• Knowledge of hardware architectures (GPU, network, storage) is a plus.
• Programming skills:
  • Python, Bash.
  • PyTorch, TensorFlow.

Location & Travel

The position can be based on several Eviden HPC sites in France: Grenoble (38), Les Clayes-sous-bois (78), Bruyères-le-Châtel (91), Bordeaux (33), Rennes (35), Toulouse (31) or Montpellier (34). Other European sites are possible. Short trips within France and Europe may be required.

Here, your ideas, curiosity, and technical excellence directly shape the next era of advanced computing – unlocking enterprise value, accelerating scientific progress and driving positive impact for society.