CareRouter

You don’t have to know where to start. We’ll take it from here

Comment

The Problem

Kingston officially declared a mental health and addictions crisis 3 years ago, with city officials reporting that conditions have continued to worsen rather than improve. Often linked to rising rates of homelessness, poverty, and substance use, leaving many people unsure where to turn for help. As of early 2024, city officials confirmed that the situation remains an ongoing crisis. - https://www.kingstonist.com/news/city-of-kingston-declares-mental-health-and-addictions-crisis/ While none of us have personally experienced a mental health emergency, our time in university has made us understand the importance of accessible and timely mental health support. When stress, anxiety, or emotional baggage overwhelms you, knowing where to start can be just as challenging as asking for help itself.

Our Solution

To address this issue we created CareRouter. It is an AI-powered triage platform that listens to your story, understands your constraints, and routes you to the exact right care in seconds. Rather than making the generic mental health app that offers unverified advice, CareRouter connects users to trusted, professional sources of support: including crisis helplines, reputable online resources, and real, bookable local services. Through adaptive questioning and real time analysis of responses, CareRouter assesses the level of emotional distress and deals with it accordingly. It gives you the right sources needed for you through helplines, websites and real bookable locations. CareRouter goes beyond guidance by offering an automated booking system, allowing users to schedule appointments with the right service at a time that works best for them with the single click of a button! We are not building a conversational chatbot for novelty; we are creating a practical, community focused solution that improves access to mental health support for everyone in Kingston.

How we built it

The application was built using Next.js for the frontend and Python for the backend, providing a fast and scalable foundation. Then the system uses ElevenLabs for voice recognition and text to speech conversations from our AI model with a voice-based check-in. To intelligently interpret user input, we integrated the Gemini API to classify and structure responses into meaningful JSON data. These structured responses form the core of CareRouter’s decision-making process, allowing the platform to adapt its guidance based on the user’s emotional state and needs. These responses from the user are the building blocks of their great mental health journey. Finally, we integrated the Google Maps API to support location-aware referrals, enabling users to discover relevant mental health resources and services near them.

Challenges we ran into

One of our biggest challenges was designing a system that could respond to emotional distress responsibly and ethically. Balancing real-time AI analysis with user safety required careful decisions around escalation, language, and user control. Mental health support is highly contextual, and translating nuanced human experiences into automated flows required careful design to avoid oversimplification or inappropriate responses. From a technical side, our biggest challenges were the linking of api keys between the various attributes of both the frontend and backend as they were linked together. There were a lot of naming conflicts and rules that needed to be set globally for every part to co-exist and ultimately bring about a smooth experience for the user. Generative AI processing, and location-based resource routing was also a big deal, as we needed to collect the current location of users to better their findings of ways to get help.

Accomplishments that we're proud of

We’re proud of building a fully functional, end-to-end voice-first mental health triage platform in a short amount of time. CareRouter successfully analyzes real-time voice input, adapts its responses based on user distress levels, and routes users to appropriate, location-aware support. The integration enabled real-time voice check ins and trauma-informed text-to-speech responses, creating a more human and supportive interaction. Achieving smooth, low latency voice interactions within a complex, AI-driven system was a major technical accomplishment, especially within the limited timeframe of the hackathon.

What we learned

Through this project, we learned the importance of human-centered and ethical AI design, particularly in sensitive domains like mental health. We gained experience integrating generative AI systems while maintaining guardrails that prioritize safety, transparency, and user agency. Technically, we strengthened our skills in building modular, API-driven systems and coordinating frontend and backend development under tight time constraints. We also learned how crucial clear communication and iterative testing are when developing user-facing experiences that rely on real-time interaction.

Future of CareRouter

Looking ahead, we envision CareRouter expanding beyond mental health to support more communities, with additional languages, broader resource networks, and deeper personalization. Future iterations could include preventative check-ins, expanded booking integrations, and partnerships with local healthcare providers and universities. Ultimately, our goal is for CareRouter to become a trusted first step, a tool that helps people move from uncertainty to support, quickly and compassionately, when it matters most in various branches of life.

Built With

Share this project:

Updates