This paper addresses the critical challenge of semantic drift and latency in multilingual telemedicine through a novel hybrid edge-cloud architecture. While teleconsultation demand increases globally, existing cross-lingual tools often fail to accurately preserve complex clinical intent or meet real-time performance requirements. We propose a framework that couples edge-based automatic speech recognition using OpenAI Whisper-v4 with cloud-based large language model reasoning via Gemini 3.1. Central to this architecture is a semantic alignment layer that maps natural language inputs to standardized medical ontologies, such as SNOMED-CT and ICD-11. Such hybrid approaches are essential for maintaining diagnostic integrity in distributed digital health environments . Evaluation using a simulated multilingual dataset demonstrates that the proposed system achieves a semantic fidelity score of 0.95 and an average end-to-end latency of 200ms. These results indicate that ontology-aware reasoning significantly outperforms standalone translation pipelines in preserving clinical intent. This research advances the methodology for real-time medical translation by addressing phonetic substitution errors and colloquial symptom mapping, which are common points of failure in generalpurpose models . The proposed framework provides a robust foundation for scalable, high-fidelity multilingual teleconsultation..