Attention-Deficit/Hyperactivity Disorder (ADHD) affects about 8% of children and often continues into adulthood. While many symptoms of ADHD are observable, its assessment frequently relies on subjective reports, leading to potential inconsistencies and biases. I aim to develop the first automated system for assessing ADHD in children through video recordings of parent-child interactions. By integrating advanced machine learning techniques to analyze visual, vocal, and verbal behaviors, I will provide an interpretable evaluation. This approach will explore gender-related differences in key behaviors and enhancing the understanding and assessment of ADHD while laying groundwork for future applications in evaluating other neurodevelopmental disorders.