Philippe Schwaller  Philippe Schwaller photo         

contact information

Predoctoral Researcher
Zurich Research Laboratory, Zurich, Switzerland
  +41dash44dash724dash86dash91

links


profile


I received my Bachelor’s and Master’s degrees from EPFL (École Polytechnique Fédérale de Lausanne, Switzerland) and I spent the last year of my Bachelor’s studies as an ERASMUS student at the University of Manchester. During my Master’s research, I had the opportunity to be a member of Prof. Nicola Marzari’s laboratory. There, I started the extraction of layered materials from inorganic crystalline compounds databases and calculated the binding energy of the layers, as well as the elastic constants of the parent crystals, with density functional theory. I was awarded the “Best Academic Master Thesis 2016” prize in Materials Science & Engineering at EPFL. After graduating in October 2016, I completed the last months of my civil service in the Functional Polymers laboratory at EMPA.

Since March 2017, I have been working for IBM Research – Europe in the Zurich lab, first in the Cognitive Computing& Industry Solutions and then in the Future of Computing department. My main focus is on data-driven discovery and synthesis of novel molecules and materials. During this period, I also studied for one year at the University of Cambridge in the department of Physics and I am currently pursuing my PhD studies in the Reymond group at the University of Bern. 

One of my publications is the Molecular Transformer for Chemical Reaction Prediction and Uncertainty Estimation presented at the “Machine Learning for Molecules and Materials” workshop (NeurIPS 2018) and published in ACS Central Science. A trained Molecular Transformer model can be used for free through our chemical reaction prediction platform: IBM RXN for Chemistry. Since the platform was launched at the ACS meeting in Boston (August 2018), more than 6500 users have joined the platform and generated over 180k reaction predictions. Recently, we added a retrosynthesis planning tool that proposes step-by-step recipes for a given target molecule, starting from commercially available precursors. 

Recently, we discovered that transformer models capture how atom rearrange in chemical reactions when trained without supervision. We used this information to build an attention-guided atom-mapper. A demo can be found on http://rxnmapper.ai.