Damien Querlioz

Short Bio

Damien Querlioz is a CNRS Research Director at the Centre de Nanosciences et de Nanotechnologies of Université Paris-Saclay and CNRS. He was first trained at Ecole Normale Supérieure Paris as a physicist and at Ecole Supérieure d'Electricité as a microelectronics engineer. He received the Ph.D. degree from Université Paris-Sud in 2009. He was then a Postdoctoral Scholar at Stanford University and at the Commissariat a l'Energie Atomique. 

Damien Querlioz develops new concepts in nanoelectronics relying on bio-inspiration. He believes that nanoelectronics can allow inventing highly energy-efficient forms of memory-centric computing. He investigates stochastic approaches,  the use of spintronics devices for bioinspired systems, and  the connection between bioinspired memories and Bayesian inference. His research interests have also included the physics of advanced nanodevices. He has developed the Wigner Monte Carlo approach to simulate and understand quantum transport in nanometer-scale devices. 

In 2016, Damien Querlioz received the Habilitation degree (HDR). In 2017, he cofounded the interdisciplinary INTEGNANO research group,  with amazing colleagues working around an integrative and multidisciplinary approach to the development of novel charge and spin-based devices, which he led until in 2023. He works with fantastic students and postdocs.  

Damien Querlioz has coauthored 9 book chapters, more than 150 journal articles and conference proceedings and given more than 100 invited talks at national and international workshops and conferences.  He has also coauthored the book "The Wigner Monte-Carlo Method for Nanoelectronic Devices" (London: ISTE; Hoboken: Wiley, 2010) with Philippe Dollfus.  In 2017, he received a CNRS Bronze medal.  He has also been a co-recipient of the 2017 IEEE Guillemin-Cauer Best Paper Award and of the 2018 IEEE Biomedical Circuits and Systems Best Paper Award.

The research of Damien Querlioz has been funded by the European Research Council, the Seventh Framework Programme of the European Union, Horizon 2020, Horizon Europe, Agence Nationale de la Recherche, Région Ile-de-France/DIM NANO-K and Ministère de l'écologie, du développement durable et de l'énergie. Starting 2022, he has been leading the BioElectronPhoton project of PEPR Electronique. 

Research

The progress of Artificial Intelligence is transforming the landscape of computing. But this comes at a considerable cost: energy. Computers and graphics cards were not developed with cognitive algorithms in mind. Brains, by contrast, have evolved to address specifically these tasks under extremely severe energy constraints. Therefore, I believe that brain-inspired paradigms can allow us to perform Artificial Intelligence in a much more energy efficient manner. And turns out, emerging nanotechnologies allow us to do precisely that!

My research is organized along two fundamental insights:

 1.  Brains do not separate computation and memory

Computers and graphics card rely on evolutions of the von Neumann paradigm: computation units and memory are separated, conceptually and physically. When they perform artificial intelligence tasks, more energy is used to move data between computation units and memory than to actually compute! By contrast brains limit considerably this data movement by using synapses as long term memory, which is located physically right where it needs to be used. This is possible because the “algorithms” used by the brain have a natural topology that do not require data movement.

Reproducing this idea with electronics is however complicated: we need to use static random access memory, which is very large, and volatile. Interestingly, nanotechnologies are now providing new classes of resistive memories that resemble brains’ synapses and can be put right in the middle of CMOS logic! In my research, I explore how we can use these devices in circuits specialized for cognitive algorithms and where computation and memory are entirely merged. More precisely, I am working on different flavors of neural networks and of Bayesian inference.

Look at my two projects in this direction: Binarized Neural Networks and Bayesian Nanoelectronics.

2. Brains compute in an approximate fashion, by exploiting device physics

Computers and graphics card compute in an extremely accurate manner using Floating Point arithmetics. To achieve perfect determinism, the physics of their basic devices is highly abstracted: an object as complex as a transistor is used as a switch.

Turns out, the precision of computers is not needed for cognitive algorithms. The brain, by contrast computes in an approximate fashion, using extensively the physics of its nanodevices like ion channels and synapses.

Current development in nanoelectronics provide us with nanodevices that have very rich physics, which we would love to use for energy-efficient computation. They would be too imprecise for conventional algorithms, but they can be perfect for cognitive algorithms!

Look at the detailed description of this research: Computing With Physics.

All this work is highly interdisciplinary and can only happen within a context of strong collaboration, and I am happy to work with an ensemble of fantastic collaborators!

Recent external collaborations: Julie Grollier at the CNRS/Thales lab (Palaiseau), Fabien Alibart at IEMN (Lille), Jacques Droulez and Pierre Bessière at ISIR (Paris), Emmanuel Mazer at LIG (Grenoble), Raphael Laurent at La Poste/ProBAYES (Grenoble), Elisa Vianello and Etienne Nowak at LETI (Grenoble), Jean-Michel Portal and Marc Bocquet at IM2NP (Marseille), Sandip Tiwari at Cornell University (USA), Manan Suri at IIT Delhi (India), Omid Kavehei at the University of Sydney (Australia), Jayasimha Atulasimha at Virginia Commonwealth University (USA). 

Contact Information

Email: damien.querlioz@c2n.upsaclay.fr 

Office Phone: (+33)170270403

Postal Address: Damien Querlioz, Centre de Nanosciences et de Nanotechnologies,  10 Boulevard Thomas Gobert, 91120 Palaiseau, France