Dental caries, periapical lesions, and impacted teeth remain among the most prevalent yet underdiagnosed oral health conditions globally, largely due to limited access to specialist radiologists and delays in manual interpretation of dental radiographs. This paper presents the design and implementation of a hybrid ML and AI-powered dental X-ray analysis and diagnosis prediction module integrated into the Neurodent web-based dental clinic management system. The system employs a two-stage architecture: a YOLOv8n model trained on dental radiographic data performs deterministic anomaly detection and localisation, producing structured bounding-box findings; these findings are then passed to Meta’s Llama 4 Scout (17B) multimodal large language model, accessed via the Groq inference API, which synthesises higher-level clinical diagnoses, urgency classifications, and patient-facing summaries. This separation of visual detection from clinical-language reasoning improves traceability, eliminates hallucinated bounding boxes, and grounds all diagnoses in explicitly detected radiographic evidence. The system delivers real-time pre-screening without requiring specialist radiologists, making it suitable for telemedicine and resource-constrained dental settings