BeHeal | Devpost

Inspiration

Our inspiration for BeHeal emerged from recognizing a significant gap in mental health support for communities affected by natural disasters. Bryson, a psychology major and member of our project team, highlighted how these events often leave lasting emotional impacts that aren’t fully addressed by traditional disaster relief efforts. Together, with our interdisciplinary team’s skills in psychology, machine learning, and web development, we set out to create a tool that empowers volunteers with realistic training to deliver compassionate mental health first aid.

What BeHeal Does

Using IBM’s WatsonX, BeHeal simulates conversations with AI-generated personas representing individuals impacted by various disasters. Volunteers practice responses and receive feedback powered by a Retrieval-Augmented Generation (RAG) system, which cites sources alongside its advice. This feature ensures volunteers can verify the guidance they receive, building confidence in delivering accurate, compassionate mental health support.

How We Built It

To build BeHeal, our team combined diverse expertise to create a comprehensive, AI-driven platform. Aaron led backend development, ensuring seamless integration between WatsonX models and the user interface for smooth, responsive interactions. Jen focused on prompt engineering, carefully crafting language that allows the AI to simulate realistic survivor conversations and deliver meaningful feedback for volunteers. Danush implemented the WatsonX models, particularly configuring the Retrieval-Augmented Generation (RAG) system to provide reliable, sourced guidance. Will handled the frontend, designing an intuitive, user-friendly interface that makes the platform accessible for volunteers. Bryson contributed to prompt engineering and took on project management, coordinating tasks and timelines to keep us on track. Together, we built a platform that not only simulates lifelike interactions but also provides trustworthy, actionable insights to equip volunteers for effective mental health support.

Challenges We Ran Into

Creating BeHeal was both rewarding and challenging, particularly when it came to integrating psychology with the project statement. Through extensive brainstorming and dedicated effort, we established a clear mission that addressed both mental health needs and the broader goal of addressing the impacts of natural disasters. Once this direction was set, prompt engineering became the next obstacle—ensuring the AI could simulate realistic interactions and provide appropriate feedback. However, after using WatsonX’s Prompt Studio, we successfully developed an effective prompt that fulfilled our objectives.

Accomplishments We're Proud Of

It took a long brainstorming session to come up with our idea for BeHeal, and we are really proud of how we took ideas from multiple team members and combined them into something that can be used to truly make a difference in the lives of victims of natural disasters and their communities. Another specific thing that we are proud of is how we leveraged Watsonx to cite sources based on specific articles and PDFs we provided. We believe that this adds another element of usefulness to the user as they can further study the psychology behind mental health after traumatic events.

What We Learned

Through building BeHeal, we gained valuable insights into teamwork, time management, and the power of AI tools. For many of us, this was our first hackathon, so working under a tight deadline taught us how to focus, prioritize tasks, and adapt quickly to new challenges. Collaboration was also key; we learned how to leverage each other’s strengths and communicate effectively to bring our idea to life. Additionally, working with WatsonX revealed its vast capabilities, and we were impressed by how much we could accomplish with its tools for natural language processing and feedback generation.

What's Next For BeHeal

Moving forward, BeHeal will support more natural disaster scenarios and a broader range of survivor experiences. Currently, the training tool allows volunteers to speak to simulated survivors of hurricanes, earthquakes, and wildfires. Future updates hope to include floods and tornadoes, expanding the awareness and preparedness volunteers gain when interacting with a variety of scenarios.