Mental health disparities in minority populations are not just a health issue but a complex interplay of cultural, economic, and systemic factors. These groups often encounter obstacles such as limited access to care, cultural stigma, and a lack of culturally competent resources, leading to underdiagnosed and undertreated conditions. Beyond traditional clinical settings, novel data analysis methods can be used to mitigate these challenges. Due to the continued growth of online social networks, there has been a plethora of information available, and this digital footprint is a significant data source to gain insights into solving the above challenge. Our research focuses on detection of mental health problems among underrepresented populations using predictive analytics on online social network data. By employing machine learning models and natural language processing, this research identifies behavioral markers and potential barriers to accessing mental health care. The findings emphasize the potential for AI-driven tools in reducing mental health disparities by facilitating targeted interventions in at-risk communities. This research further discusses ethical concerns related to AI-driven health assessments and highlights challenges for their responsible deployment.