Antoine Ottiger

– Developed a surrogate model (PyTorch) to predict eigenmodes of mechanical structures, replacing finite – element simulations (30min → <1s) with an average relative error of 1.4%. – Leveraged the model for identifying unknown physical parameters on a real system.

– Fine-tuned a multilingual translation model (PyTorch/Hugging Face) on a corpus of industrial technical – documentation, enabling learning of specialized industry vocabulary. – Developed a WinForms application (C#) integrating the model to translate technical documentation, – deployed to engineering teams. – Evaluated and optimized the model using BLEU/COMET metrics, achieving performance close to professional references on target languages.

– Developed a Python software that automated the design of textures for laser machining. – Implemented Cognex industrial vision systems for automatic part detection and positioning; explored alternative machine learning methods for annotation-free localization. – Developed technical demonstrators for laser machining solutions: CAD (Rhinoceros / Grasshopper 3D), laser parameter optimization, 3D metrology, 5-axis machines, forming, texturing, and nano/femtosecond laser technologies.