This paper studies the relationship between migration and the productivity of high-skilled workers, as captured by inventors of international patent applications. Using machine learning techniques to uniquely identify inventors across patents, we are able to track the migration patterns of nearly one million individual inventors across countries. Migrant inventors account for more than nine percent of inventors in our sample. The econometric analysis seeks to explain the recurring finding in the literature that migrant inventors are more productive than non-migrant inventors. We find evidence that migrant inventors become about twenty-three percent more productive after having migrated. The disambiguated inventor data are openly available.