Code similarity systems can power several tools (e.g. code recommendation system, automated bug detection, automated performance engineering) that improve developer productivity and increase the security, stability, and performance of the software applications they build. However, existing structural representations of code, required to build code recommendation systems, have limitations that inhibit their practical application. To resolve these limitations, Ye et al. present Machine Inferred Code Similarity (MISIM), an end-to-end code similarity system that leverages a context-aware semantic structure designed to extract semantic meaning from code syntax; and a deep learning-based similarity scoring algorithm, which can be implemented with different architectures. MISIM outperforms existing code similarity systems by 1.5 – 43.4x.