Cancer and diabetes are complex diseases that have proven difficult to treat in the clinics. Until recently, most efforts have focused on hitting individual therapeutic targets cleanly. Unfortunately, success rates in clinical trials have been low: in the case of cancer, success rates for individual drugs have commonly been in single digits. This suggests that our models are poorly predictive of useful therapeutics. One issue is disease complexity. Another is the importance of modeling disease in the context of the whole animal. I will discuss our efforts using Drosophila to build complex cancer and diabetes models in the context of the whole animal. We utilize genomic sequencing data to create multi-hit fly models, either of disease populations or of individual patients. We then use automated systems to screen libraries for useful drugs. In addition, I will describe how we combine Drosophila genetics plus medicinal chemistry to create new generation candidate therapeutics.