This U.S.–Swiss project tackles a major barrier in the development of silicon photonic chips — a key technology for faster, energy-efficient data movement in computing, communications, and sensing. Current design methods are slow, expensive, and hard to scale. This project introduces Fusion Network (FusNet), a novel approach that combines simulation and design optimization into a single, fast process. By using advanced computing hardware and AI-guided techniques, FusNet drastically reduces the time and effort required to create complex photonic circuits. Beyond technical innovation, the project builds a global training pipeline in semiconductor design, featuring U.S.–Swiss student exchanges, curriculum integration, and industry-aligned programs with Cadence Design Systems. The results — including software tools and design methods — will be shared openly to benefit the broader research and engineering communities. FusNet is a hardware-accelerated framework that unifies electromagnetic simulation and optimization in a real-time loop, enabling one-shot photonic device design. It leverages structural similarities between forward and adjoint methods to merge the simulation and optimization processes, improving efficiency while maintaining accuracy. FusNet’s streaming architecture is implemented on high-throughput dataflow hardware, such as CGRAs and FPGAs, overcoming traditional bottlenecks in memory access and parallelization. The project includes algorithmic innovation, scala