Design an architecture for an advanced cross-lingual information extraction system that works for at least 10 languages. Include: 1) Model architecture considerations for handling languages with different morphological structures. 2) Strategies for effective transfer learning from high-resource to low-resource languages. 3) Methods for handling code-switching and mixed-language documents. 4) Approaches for extracting complex nested entities and relations. 5) Techniques for handling domain-specific terminology across languages. 6) Evaluation framework that accounts for linguistic nuances and annotation challenges. 7) Strategies for dealing with ambiguity and context-dependent interpretation. 8) Optimization for low-latency inference in production environments.