Precise nanofabrication represents a critical challenge to developing semiconductor quantum-dot qubits for practical quantum computation. Here, we design and train a convolutional neural network to interpret scanning electron micrographs and quantify qualitative features affecting device functionality. The high-throughput strategy is exemplified by optimizing a model lithographic process within a five-dimensional design space and by demonstrating a robust approach to address lithographic proximity effects. The results emphasize the benefits of machine learning for developing stable processes, shortening development cycles, and enforcing quality control during qubit fabrication.