In this talk, Silvio Amir will present his ongoing work towards addressing these issues. First, he will discuss novel methods and datasets to identify medical claims on social media and retrieve peer-reviewed literature to support/refute said claims. He will then show how to reduce manual efforts in deploying these applications in other domains by eliciting annotations from large language models. Second, he will discuss the opportunities and challenges of incorporating images and personal attributes in models for longitudinal social media data. Then, he will present a case-study on the application of personalized multimodal models to estimate the prevalence of various mental health conditions such as depression, anxiety, and PTSD from social media.