Bestie by Warm

AI Based Smart Assessment based on daily conversations through Chatbot for personalizing mental health services such as On-Demand Therapy, Content Provision, and Volunteer-Based Suicide Prevention.

Comment

Inspiration

Mental health is a pressing issue in Indonesia, with over 19 million adults experiencing emotional disorders and 12 million affected by depression. Sadly, 63% citizens cannot barely access the service, with only 4 clinicians for 1 million people. Various frameworks and methods have been proposed to address the lack of resources and access barriers, starting by remotely assessing the source of illness. However, conventional assessment methods are time-consuming, tedious, complex, and undemocratic. But, even though similar problems happen there, a company from Thailand, Sati, try to solve this by using volunteer-based suicide prevention. Such an inspiration for us in the middle of limitation, they really trying to take the part as intervention actor.

What it does

Bestie is a health issues assessment tool that utilize conversational AI. Bestie revolutionizes how we do health issues assessment, traditionally by doing self-assessment check list into light conversation with AI chatbot. The conversation data then compiled into concise statements that integrates seamlessly into other features.

  • Revolutionizes mental health issues assessment
  • Utilizes conversational AI to summarize into concise meaningful statements, then further classified into common mental health problems

How we built it

Initially, we diligently studied the syntax of OpenAI and the methodology of implementing a chatbot using its API. Subsequently, we conducted extensive research to identify models available on the internet capable of learning sentiment analysis. Armed with this knowledge, we proceeded to design an intuitive user interface that resonates with our users.

As development progressed, we devoted ourselves to coding the interface while simultaneously establishing the backend operations to form a robust serverless network. This approach allowed us to streamline the overall process, ensuring seamless integration between the front-end and back-end components.

In essence, our endeavor involved a harmonious amalgamation of technical expertise and creative ingenuity to bring forth a well-crafted, user-friendly chatbot assessment platform.

Stack Used

  • Backend: Python3 with OpenAI's GPT.
  • Frontend: Dart
  • Deployment: Deployed as a Serverless Architecture using AWS Lambda

Development Process

We started by brainstorming ideas. We refine our ideas overnight and started prototyping and designing in Figma soon after.

In the morning we finalized the design idea. In parallel, we did research upon the infrastructure and tech stacks that we'd like to utilize. In the afternoon we started hacking the frontend, backend, as well as refining the design interaction details. Following that, we did rigorous testing and bashing some bugs we found along the way.

Challenges we ran into

We encountered challenges while implementing the serverless system due to the frequent errors/bugs in the API provided by OpenAI. Moreover, displaying recommendation buttons proved to be a difficult task as we needed to make them dynamic and conditionally appropriate. Therefore, it was essential to employ the right methods to overcome these obstacles.

Accomplishments that we're proud of

We are proud to have implemented a serverless system in building Bestie, the chatbot assessment. Additionally, we take pride in designing a bot that fulfills its responsibility in conducting the Brief System Inventory assessment on patients according to the provided guidelines. We are also pleased that the bot we designed can draw conclusions about the patients' moods based on their responses.

What we learned

GPT (generative pre-trained transformers) has advanced rapidly to the point that it can be publicly accessible using API (but still regulated and limited). The usage is unlimited and only bounded to our creativity itself.

What's next for Bestie

In the future, Bestie is expected to improve its accuracy in assessing someone's mood. Bestie is expected to perform data training to discover the appropriate assessment model, ensuring that the parameters used are clinically tested and consistent. Not only that, Bestie is also expected to enhance accuracy and customization in the services provided to Warm application users who seek to improve their mental well-being.

Built With

Share this project:

Updates