In this talk I will first describe a novel methodology for the fast evaluation of donor/acceptor systems for photovoltaics. The new approach, up to 100 times faster than conventional optimization protocols, is based on the use of Raman to quantify the local thickness and composition in samples with lateral gradients on parameters of interest. Raman images are combined with photocurrent images (LBIC) to identify the optimum conditions. We demonstrate the potential of the methodology optimizing three systems PCDTBT:PC70BM, PTB7-Th:PC70BM and PffBT4T-2OD:PC70BM, obtaining efficiencies circa 6%, 8% and 10%, respectively, using less than 50 mg of each polymer in the process. I will show that this method can be used also to analyze blends containing non-fullerene acceptors and ternary systems and it can be extended for the case of evaporated bilayer solar cells by using moving shadow masks as well as polymer:polymer processed through microfluidic chips dispensers. Finally, I will describe our first attempts to use these large datasets (>25.000 points per material system) as input in machine learning algorithms and what we could learn from this exercise.