Inspiration
The inspiration for PatSafe came from the need to improve patient care and communication post-discharge. After being discharged, patients often struggle with tracking their recovery progress and medication adherence. Similarly, doctors face challenges in monitoring their patients remotely and ensuring they are on the right track. PatSafe was created to bridge this gap by providing an easy-to-use platform that allows seamless communication between doctors and patients, ensuring better recovery outcomes and more informed healthcare decisions.
What it does
PatSafe is a web-based application that connects discharged patients with their doctors for post-discharge care. It features a doctor’s dashboard where they can update patient information, provide medication instructions, and track patient progress. Patients can report their symptoms, recovery status, and medication adherence, which is then updated in real time for their doctor to review. Additionally, PatSafe includes an AI-powered chatbot that allows patients to ask medical questions and report symptoms, helping doctors get insights into the patient’s condition remotely. The chatbot also aids in tracking medication adherence and offering information for diagnoses and prescriptions.
How we built it
PatSafe was built using Next.js and React.js for the frontend, while the backend is powered by MongoDB and Clerk for authentication. The platform utilizes APIs for real-time updates and features such as medication tracking and symptom reporting. We integrated a chatbot powered by Hugging Face models for natural language processing, allowing users to interact conversationally and gather helpful medical information.
Challenges we ran into
One of the biggest challenges we faced was ensuring smooth integration between the doctor’s dashboard and the patient’s input. Additionally, our original plan for a chatbot was a fine-tuned model from Hugging Face on a Databricks compute cluster. The cluster never started no matter the configuration, always running into Azure errors when starting and the AWS version having errors creating a workspace, and fine-tuning using MosaicML ended up not working out. We settled for a Hugging Face model called using LangChain, hosted on Friendli.
Accomplishments that we're proud of
We’re proud of the successful integration of the real-time doctor-patient communication system that allows immediate updates between both parties. The chatbot’s ability to provide helpful medical information, track symptoms, and assist in medication adherence was a key accomplishment. Additionally, the doctor’s dashboard was designed intuitively to help healthcare providers make informed decisions based on the data patients provide.
What we learned
We gained deep insights into how AI can be used in healthcare to improve communication and decision-making, and more exposure to Databricks tools. Additionally, we learned the importance of making systems scalable and future-proof while maintaining a simple and intuitive user experience.
What's next for PatSafe
Fine-tuning a model using MosaicML and having it on a Databricks cluster with MLflow to continue its learning would be a great next step, as well as verification for doctors and patients, real-time updates using Websockets for the dashboard, function calling from the chatbot, and a custom AI-generated daily survey from the doctor's data for the patients to do.
Log in or sign up for Devpost to join the conversation.