We propose a new, systematic approach to engineer the substrate specificity of enzymatic catalysis, a long-standing problem with no general solution, using S1A proteases as a model system. Structural biology has identified mutations that are often necessary but not sufficient to convert specificity. In vitro evolution has solved this problem in a few cases but at the cost of inefficient exploration of sequence space, for lack of a robust rationale to focus mutations and to choose favorable starting points. Our new general approach starts with measuring the catalytic activity of thousands of proteases across the enzyme family towards several substrates, in a cost/time-effective high throughput experiment combining droplet microfluidics and next generation DNA sequencing. Next, we will extract from the data sequences of ‘generalist’ enzymes with broad specificity profiles as favorable starting points for the in vitro evolution of specificity, and focus mutations at specificity-determining residues we have identified via statistical data analysis. From the data we will also infer a generative statistical model of the relation between enzyme sequence and specificity profile, which will predict the specificity of uncharacterized proteases, mutations in known proteases to engineer their specificity, and ultimately sequences of synthetic proteases with tailored specificity profiles. Beyond solving a fundamental problem, our work has direct applications in peptide sequencing and mass spectrometry analysis, and its potential applications extend to green chemistry, biotechnology, the agro- and the pharmaceutical industry.