Medical resources are often limited, particularly in densely populated areas where the demand for hospital diagnoses is high, leading to long waiting times for appointments. This situation can be frustrating for patients and contributes to inefficiencies in healthcare delivery. If patients could effectively describe their symptoms and receive preliminary diagnoses promptly, it would alleviate pressure on healthcare systems and empower patients to manage their conditions better. This study leverages the open-source Meta AI language model LLaMA2 to address this challenge. First, text de-duplication is performed using TF-IDF vectorization and cosine similarity filtering to ensure data quality. The model is then fine-tuned using the Hugging Face PEFT library, employing LoRa fine-tuning and quantization techniques on an NVIDIA 3090 device, enabling it to make initial assessments based on patient symptom descriptions. Finally, the model's performance is evaluated by attempting the United States Medical Licensing Examination (USMLE), assessing its ability to provide medical insights and support healthcare professionals effectively.