Unified Interactive Frameworks for Modular Coding and Circuit Design

This thesis presents a unified framework for scalable, modular, and interactive development across software and hardware domains. We introduce BHDL, a declarative embedded language for PCB design that enables concise, hierarchical schematics and layouts within a Jupyter-powered environment. Building on this, we propose a deep reinforcement learning-guided Monte Carlo Tree Search approach for circuit routing, achieving flexible, high-performance automation adaptable to diverse design constraints. To overcome the limitations of traditional notebooks, CodePod extends Jupyter with hierarchical scoping and language-agnostic modularity, supporting large-scale, multi-language projects. Finally, Kernel-FFI enables seamless cross-language function calls and object referencing in interactive workflows, eliminating boilerplate and supporting object-oriented patterns and recursive remote function calls. Together, these contributions establish a cohesive ecosystem for interactive, modular, and scalable development in both coding and circuit design.

Committee: Jin Tian (co-major professor), Chris Quinn (co-major professor), Pavan Aduri, Wensheng Zhang, and Forrest Bao