Professor Emre Neftci Visits B2SLab

projects
B2SLab welcomed Professor Emre Neftci for a seminar on In-Memory Computing for Fast and Energy-Efficient Generative AI, exploring new approaches to making generative AI faster and more energy-efficient.
Author

Jordi Fonollosa

Published

October 23, 2026

Modified

November 19, 2026

We were pleased to receive the visit of Professor Emre Neftci, Director at Forschungszentrum Jülich and Professor at RWTH Aachen University. He presented a talk entitled “In-Memory Computing for Fast and Energy-Efficient Generative Models”.

During the seminar, he explored how analog in-memory computing can overcome the limitations of speed and energy efficiency of processing large sequences, particularly when running generative models on edge devices. Emre presented recent advances in hardware-algorithm co-design for transformer-based architectures, focusing on self-attention mechanisms, memristive crossbar arrays and hardware-aware optimisation.

After the talk, we had a nice discussion on the future of AI and its applications. Emre, thanks for coming and visiting us!