Machine learning (ML) has become a widely-adopted tool in materials engineering and research. However, off-the-shelf approaches often fail to solve problems in materials science. I will discuss domain-specific challenges that arise when applying ML to materials.