Boxing is a high-intensity combat sport with asymmetric motions, explosive actions, and a high danger of overtraining and injury. This means that rigorous monitoring is necessary for athlete heath. This research presents an AI-driven framework that amalgamates wearable inertial measurement sensors (Neuron32), machine learning models, and a multi-objective recommender system to oversee performance, forecast fatigue, assess training load, and avert injuries in both professional and amateur boxers. We kept an eye on 15 seasoned athletes and 97 amateur athletes, looking at things like exhaustion, training load, and injuries. We created predictive models for classifying fatigue and estimating load, and we used their results to build a recommender system that finds a compromise between improving performance and lowering risk. The results showed that the models were quite good at predicting outcomes. The fatigue classification had an AUC of 0.91, and the load prediction had a R² of 0.89. The recommender system also cut down on fatigue-related overload episodes by 21.3% and injuries by 13.8%. These results indicate that AI can provide objective, tailored insights for boxing training, facilitating both performance improvement and healthfocused approaches. In conclusion, this architecture shows how AI-enabled monitoring systems could improve precision sports medicine in boxing and be used as a paradigm for other combat and endurance sports.