Built by Physicians. Powered by AI.

The term “Pendium” is derived from the word “compendium”, a comprehensive and extensive collection of medical knowledge and information. Historically, doctors used to rely on physical compendiums, which were large textbooks containing pharmaceutical and specialty information. These compendiums were cumbersome to carry and navigate. Pendium Health embraces the essence of this concept by modernizing and streamlining how doctors access medical information using cutting-edge AI technology. Our platform aims to provide a digital and intelligent alternative to the traditional compendium, enabling healthcare professionals to access a vast repository of medical insights seamlessly.

At Pendium Health, we leverage the power of Artificial Intelligence and Large Language Models to revolutionize the way doctors access medical information. Our AI-driven platform intelligently organizes and categorizes a wealth of medical peer reviewed literature. It offers contextual recommendations, rapid search capabilities, and real-time updates, enabling healthcare professionals to make informed decisions efficiently. By harnessing AI, we empower doctors to navigate through complex medical knowledge swiftly and accurately, ultimately enhancing patient care and medical practices.

HippoAI is revolutionizing the way doctors search for information by streamlining clinical queries using the power of Large Language Models and semantic search. Hippo can handle most clinical questions and provide evidence-based answer with citations and links for further reading.

Semantic search is a search algorithm that goes beyond keyword matching. It utilizes natural language processing (NLP) and machine learning to understand the context and meaning behind search queries. For example, say a physician wants treatment information for hypothyroidism given specific lab values and patient demographics. In a traditional database, the physician may look up ‘Hypothyroidism’ and find the relevant article, then paragraph or table that has the answer. With semantic search, the physician can simply ask ‘What are the treatment recommendations for a 54F with hypothyroidism and a TSH of 10’ and our algorithm will fetch the exact paragraph from the appropriate guideline to answer this question.

HippoAI works by utilizing a combination of semantic search retrieval and Large Language Model text generation. First, we search our clinical guideline database for relevant passages from the medical literature based on a clinical question. Then, a language model reads the retrieved information and generates a response. If a question does not have a direct source in our database, Hippo is trained to respond with “I don’t know,” ensuring answers are only generated from available source text. 

Generative AI refers to models capable of producing text or images. ChatGPT is an example of a highly capable language model. These models are trained on large amounts of natural language data, enabling them to generate human-like responses to given inputs. Large language models, or LLMs, such as GPT-3.5, can efficiently analyze extensive clinical data and provide relevant information to aid medical professionals in making better decisions for patient care.

To design an effective prompt for Hippo, it is important to be specific and clear about the information you are seeking. Provide context and relevant keywords to guide Hippo’s search for relevant paragraphs in our guideline database. If the initial answer is not what you were looking for, you can try clarifying with a follow-up in the textbox below or start a new conversation thread.

While we are in active development, Hippo is currently free to use. We value any feedback you may have.

You can ask questions that have theoretical answers in published clinical guidelines. Examples include questions related to treatment or diagnostic approaches for a medical condition, screening, counseling, etc.

Hippo is not specifically trained to handle broad medical knowledge questions or provide differential diagnoses based on symptoms, unless explicitly stated in our guideline database. While it may still provide answers to these questions, they may be less reliable, and the sources provided may not be as relevant.