SentinelPharma

The InspirationThe pharmaceutical industry moves at a pace that often leaves healthcare professionals (HCPs) overwhelmed. We realized that a doctor has, on average, less than 2 minutes to spend with a traditional Medical Representative. We were inspired to bridge this "information gap" by creating a tool that provides the same expertise as a human rep but with 24/7 availability and 100% data accuracy.What We LearnedWe dove deep into the world of pharmaceutical compliance and drug reimbursement. We learned that it isn’t just about what a drug does, but how a patient can afford it. Understanding the complexity of tiered pricing and insurance coverage showed us that AI must be more than just a chatbot; it must be a multi-modal data analyst.How We Built ItWe built a pipeline focused on Retrieval-Augmented Generation (RAG) to ensure the AI never "hallucinates" drug dosages.Backend: Python and FastAPI.Vector Database: We used ChromaDB to store and index thousands of pages of drug leaflets and reimbursement policies.The Model: We utilized a Large Language Model (LLM) fine-tuned on medical corpus data.The Logic: To determine the "Reimbursement Score" ($R_s$), we implemented a simplified weighted formula:$$R_s = \sum_{i=1}^{n} (C_i \cdot W_i)$$Where $C_i$ represents the coverage factor of the insurance provider and $W_i$ is the weight of the specific patient demographic.

Built With

• adnn
• fast
• python