Inspiration
Living with asthma involves much more than tracking symptoms. Asthma follows you through everyday life—when you're deciding whether to exercise, preparing for an activity that might trigger symptoms, or simply wondering whether the way you're feeling is something you should worry about.
Our earlier Lobelia work explored asthma forecasting and personalized guidance. For BuildFest, we asked a different question: what if that technology felt less like another health dashboard and more like a companion that stayed by your side?
That idea became Lobelia Asthma Companion—an application that brings together supportive conversation, daily reflection, and access to saved support information.
Business Opportunity
Market Pain: Across 28 million U.S. asthma patients, uncontrolled exacerbations drive over 1 million emergency room visits and $50 billion in direct annual medical spending, creating a massive, concentrated financial burden for healthcare payers.
Strategic Solution: Lobelia moves asthma care upstream by deploying a risk-forecasting platform that predicts daily flare-up risks and delivers timely, contextual guidance before costly acute care is triggered.
Business Model & ROI: We see a potential B2B2C path through health plans and health systems, where better day-to-day asthma management could create value for both patients and organizations bearing the cost of acute care. Our next step is validating this model and its potential impact with patients and healthcare partners.
What It Does
Lobelia supports people with asthma across three moments: anticipating risk, navigating everyday life, and finding the right support when symptoms worsen.
Anticipate: Lobelia can surface an existing asthma risk forecast alongside relevant context such as upcoming plans and environmental conditions.
Support: The AI companion provides a friendly place to talk about your day, ask questions about your forecast, and reflect on how you're feeling. It can draw on saved preferences, upcoming plans, and recent conversation context, and users can choose a warm, calm, or direct communication style. We designed personalization around a simple idea: people don't all want support in the same way. Some may prefer reassurance, while others want concise, direct information. Lobelia uses explicit preferences and user-approved memory to shape how the companion interacts with each person.
Respond: When breathing becomes difficult, our concept provides quick access to saved asthma action-plan information and trusted contacts, with the longer-term goal of providing a structured, clinically validated escalation flow.
How We Built It
We built on the existing Lobelia codebase, adapting its React and Vite frontend, Python and FastAPI backend, and PostgreSQL database for a focused BuildFest experience.
The companion backend uses Google Gemini or Anthropic Claude to generate conversational replies. It assembles relevant information—including saved preferences, upcoming plans, an approved summary, and an existing forecast—and supplies that context alongside the user’s message.
We separated conversation from prediction: the LLM can discuss a stored forecast, but it does not calculate or change it. We also added structured response validation and fallback messages for unsuccessful generation.
On the frontend, we connected the chat interface to personalized greetings and ongoing replies, with controls for conversational tone and starting a fresh chat.
Challenges We Faced
One challenge was deciding how much of the existing application to bring into a small demo. Lobelia already had broader forecasting and AI functionality, so we had to identify which pieces supported the companion experience and keep the scope manageable.
Another was making the conversation feel attentive rather than repetitive. We refined the companion’s instructions to respond to the specific thing a person says, acknowledge feelings, and avoid turning every reply into generic advice.
Context also required careful handling. A saved plan is an intention, a forecast belongs to a particular date, and an older summary may no longer reflect what the user wants to share. We worked to preserve those distinctions and prioritize current corrections.
What We Learned
We learned that a useful AI companion depends on more than connecting a model to a chat box. Selecting the right context, handling missing information, and giving users control over what persists are central to the experience.
We also learned to distinguish conversational continuity from long-term memory. Recent messages help a conversation flow; saving information for future sessions should be a separate, deliberate choice.
Most importantly, the project helped us connect the technical foundations of Lobelia with a more approachable interaction: a place to talk about daily life while keeping relevant personal information close by.
What’s Next
Our next step is to turn the companion concept into a more complete mobile experience: finishing the character-guided experience, connecting proactive support to upcoming activities and environmental conditions, and developing the structured breathing-support flow.
Beyond implementation, our biggest priority is validation. We want to continue talking with people with asthma and clinicians to understand which features are genuinely useful, evaluate whether the companion can improve engagement with day-to-day asthma management, and clinically validate any safety-critical guidance before deployment.
We also want to explore how Lobelia could become sustainable beyond a prototype by identifying the users who benefit most and validating potential distribution models with health systems, payers, employers, and other asthma-care partners.
Log in or sign up for Devpost to join the conversation.