Harbor

Inspiration

Over half of mental health practitioners have faced burnout, an issue often associated not only with the heavy subject matter of many of their client's challenges, but also with the sheer amount of work they have to do in between sessions. Therapists will have to make detailed care plans, review homework, create goals, and field questions from clients between sessions. With AI tools serving as medical scribes for many practices, we wanted to take it a step further

What it does

Harbor does a few things. Firstly, a chatbot interface helps handle any of the client's needs in between sessions. It also allows for a secure platform for therapists to send in homework assignments. There is an agent that takes the past weeks worth of conversations and assignments to generate care plans. Finally, there is a secure video conferencing feature that records sessions and analyzes transcripts.

How we built it

We used React for the frontend and Node/express/mongo for the backend. We used gemini for the chatbot and created a RAG pipeline using conversations found in an open source database of therapist conversations (consistent with best practices according to research). We used elevenlabs for transcripts and videosdk for the conferencing. We also used gemini for the lesson plans.

Challenges we ran into

One major challenge we faced was in making the VideoSDK work. A lot of these tools we integrated were both challenging and expensive, which served as a massive bottleneck.

Accomplishments that we're proud of

We are more so proud of the fact that we were able to make something to solve this problem. A lot of these issues were brought up throughout our general research of the space, so it is exciting to see it come to fruition