Yu-Wei Tseng

My foundation is in architecture — five years at National Cheng Kung University in Taiwan, studying how physical systems serve people at scale. Circulation, structure, spatial logic, design rigor. Internships at Atelier Or and Soar Design taught me to see the world as interconnected systems and to design for human experience at every scale. It turned out to be surprisingly good preparation for building software systems that do the same thing.

The transition to software started at CtrlS Ltd., where I built serious games and AR experiences in Unity. That led me to SMU Guildhall for a Master's in Interactive Technology, where I built a custom C++ game engine (Daemon Engine) and shipped team projects — HardDriverz, Corrupted Hollow, and PlanePainter. The thread connecting it all: designing systems where humans (or agents) navigate, make decisions, and produce outcomes.

Now my focus is on agentic AI systems and distributed architectures. My thesis — AGENTS — asks whether a network of AI agents can autonomously collaborate to build a game: discussing requirements with a human, decomposing creative goals, executing in parallel via isolated git worktrees, and verifying each other's output. The system is built on the KADI ecosystem, with an orchestrator agent that bridges the human and the multi-agent pipeline through network-scoped tool discovery, recursive tool calling, and memory injection. It took 3–4 months of iteration to find the right balance between smart and rigid behavior — and that search is still ongoing.