Nowadays, benefiting from virtual and augmented reality, manufacturing sites are constantly updated through digital training platforms, becoming essential tools for efficient, secure, and scalable professional development and evolution. Latest innovations related to digital twin, intelligent computing systems, and real-time networking communication have made it possible to transform these solutions into a cutting-edge alternative to traditional methods, particularly in branches like production and manufacturing, where security, performance, and adaptability hold high priority. Considering this, an extended reality platform was conceived, deployed, and tested, guaranteeing high performance, security, and scalability through a systematic backend optimization. The architecture integrates Kubernetes orchestration, Zero Trust security via OpenZiti, and a custom STUN server for low-latency peer communication. Additionally, dynamic scalability is obtained thanks to containerized microservices, session-aware routing, and adaptive load balancing, enabling the platform to efficiently support concurrent users and distributed training scenarios in real time.