I am building a proof-of-concept that predicts a patient’s disease from entered symptoms and then powers a simple chatbot to explain the result. The core of the study will compare three large-language-model families—T4/T5, MedAlpaca and LLaMA—trained or fine-tuned on the Kaggle “Diseases and Symptoms” dataset (https://www.kaggle.com/datasets/dhivyeshrk/diseases-and-symptoms-dataset). Here is what I need: • Clean, notebook-based Python code (any libraries welcome) that trains each model, evaluates them, and exposes a lightweight chat interface. • Basic evaluation outputs: overall accuracy, class-wise precision/recall, a confusion matrix, and at least two clear performance graphs. • A concise but properly referenced literature review written as a detailed comparative analysis of similar symptom-to-disease efforts. Highlight where my approach can be improved and note any innovative angles you spot. • A short draft research report structured for eventual conference submission; I’ll share final formatting rules later, so keep the draft flexible. Keep the work modular so I can iterate easily after delivery.