The United States is experiencing an opioid crisis, with an estimated 10 million people aged 12 or older misusing opioids, and 130 deaths a day from opioid overdose (OD). The estimated economic burden of the opioid epidemic is approximately 78.5 billion dollars per year. We take an integrative big data driven approach by integrating and analyzing multi-scale data such as electronic health records, social media, social-economic data and government policies to understand the epidemic. We explore geospatial patterns and disparities at fine grained geospatial resolutions. We have developed various machine learning/deep learning based models to predict risk of patients with opioid overdose or opioid use disorder. Meanwhile, we are developing an integrated artificial intelligence pipeline to identify novel opioid analgesics.