While digital manufacturing (e.g., 3D printing) has had significant advances on the hardware and material fronts, algorithmic design for digital manufacturing remains one of the biggest barriers to its widespread adoption. Computational Design and Fabrication (CDF) is an emerging research area that tries to address this problem. Functional fabrication is an important paradigm in CDF where given the high-level function or goal, the design is computed, and physically realized by the fabrication hardware. Although the functional fabrication lends itself to powerful abstractions, we still lack a general and practical workflow. Here, I will show how machine learning methods can raise these abstractions closer to the practical level. Specifically, I will present the subject of neural invers