STAGE - Stabilité numérique des réseaux de neurones OPC sur GPU - F/H
Core
Characterize and reduce numerical differences in neural network inference results across different GPUs (A100, H100) to ensure digital consistency in high-performance computing environments.
Role type
Intern, Machine Learning Engineer (GPU numerical stability)
Builds
Optical Proximity Correction (OPC) mask correction models for semiconductor manufacturing
Domain
Semiconductor manufacturing / Machine Learning / High-performance computing
Deliverable
production ML models
Required skills
Applied mathematics, Machine Learning, Python, PyTorch, Linux, GPU programming (Nvidia)
Preferred skills
Analysis of numerical behaviors, model robustness, experimental design
Technologies
Python, PyTorch, Linux, Nvidia GPUs
Responsibilities
Analyze numerical differences in model inference across GPU hardware, identify sources of variance, explore methods to reduce or eliminate these differences, conduct experiments and compare results
Seniority
Intern, 6 months
