Designing optimal cells for electrochemical CO₂ reduction involves exploring a vast parameter space. This exploration can benefit greatly from computational screening and machine learning. In our group, this is done using a combination of approaches. In the first approach, in collaboration with Mie Andersen’s research group, we investigate the material structure on an atomistic level to study its behavior, using universal potentials as cheap surrogates for the more accurate quantum-mechanical description. In the second approach, we train machine learning models on experimental data collected in our lab to better understand the correlations between the synthesis and testing parameters and the material performance.