Description du sujet de thèse

This study aims to integrate spintronic nanodevices into hardware neural networks and train them using chip-in-the-loop techniques. The research will explore the radio-frequency (RF) properties of spintronic devices for both computation and learning, with a focus on implementing training directly in hardware. The candidate will contribute to the design and testing of RF nanodevices and circuitry for neural networks, evaluate their functional performance, and develop algorithms that leverage the underlying device physics to achieve high accuracy—despite the inherent imperfections of nanoscale components.

Activities

• Electrical characterization of RF properties of spintronic devices.
• Design of neural networks based on the nanodevices
• Design and simulation of RF circuits compatible with spintronic neural networks
• Experimental validation and characterization of the integrated systems
• Development of training algorithms and implementation in hardware
• Collaboration with a multidisciplinary research team.

Expertise

• Expertise in Python and developing neural networks training algorithms
• Expertise in experimental spintronics physics
• Expertise in the measurement and test of spintronic nanodevices
• Expertise in radio-frequency electronics

Contexte de travail

The work will be carried out at the Albert Fert Laboratory, in the "Neuromorphic Physics" team exploring the use of nanodevices and their multiple functionalities for bio-inspired computing. The team includes two permanent CNRS researchers, two Thales researchers, 4 post-docs, and 4 PhD students.

Contraintes et risques

N.A.