Developed multi-user session management for the ADK Demo within the NVIDIA/NeMo-Agent-Toolkit repository, focusing on enabling reliable concurrent user interactions. Leveraged Python and asynchronous programming to implement conversation-based session caching and per-chat session isolation, preventing cross-user data leakage and supporting smoother parallel sessions. Addressed workflow issues by refining session orchestration, which enhanced both the stability and scalability of the demo environment. The work demonstrated a strong grasp of full stack development principles, integrating caching strategies directly into the demo’s architecture. This feature improved the overall user experience by ensuring each chat session remained independent and responsive during multi-user operation.