This document presents a CAD framework called CAOS for developing reconfigurable high-performance computing architectures using FPGAs. CAOS aims to improve the usability, interactivity, and modularity of FPGA design flows. It includes frontend modules for profiling and modeling applications, optimization functions, and backend modules for generating optimized FPGA bitstreams and integrating with system runtime tools. The document also describes a case study using CAOS to implement a data science application across a host CPU, PCIe, and FPGA accelerator.