Many of the challenges we face as a society are related to finding optimal materials in areas such as energy and sustainability, healthcare, transportation, foodstuffs, .... Software-based approaches offer an unprecedented way to accelerate and improve how we design and deploy new materials. Through artificial intelligence we can leverage larger than ever experimental and theoretical datasets to find unexpected patterns and to generate novel ideas. Atomistic simulations have never been faster or more accurate and we are now enabled to routinely carry out virtual searches over large spaces of candidate materials in a predictive way.
Here, I will report how we have combined machine learning and quantum chemical simulations with experimental intuition to discover and fabricate novel practical molecular materials. I will report work in the discovery of organic light-emitting diodes for displays and lighting, organic electrolytes for electrical energy storage and dyes for novel solar cells. In addition, I will discuss recent results in the application of deep learning techniques to perform automatic chemical design through a learned continuous representation of molecules.