Menthelp

Inspiration

We saw cost of therapy combined with the declining mental health crisis. We wanted to make a solution to expand access and provide personalized, private support.

What it does

Menthelp works to help patients with their mental health concerns. It works as a therapist and will help a patient deal with their mental health through chatting with them. It provides personalized, private support.

How we built it

We used the pre-trained chatbot Llama 2 created by Meta. We then fine-tuned it using a Therapist Dataset on Google Colab, exported the weights, and used Hugging Face and Gradio to deployed it on a website using Flask and Bootsrap.

Challenges we ran into

After Charan trained the model, we had trouble loading it. We tried many ways listed in the Llama community but eventually found a way from Stack Overflow in order to reload our weights onto a model and test it in Google Colab.

Unfortunately, we cannot afford GPU on Hugging Face so we decided to just use the original Llama and Embed it in our website.

Accomplishments that we're proud of

First of all, we are all proud of being able to work together. Lucas comes from Sugar Land, Texas, Charan comes from Silicon Valley, and Aaron comes from New York. We were proud of being able to communicate from long-range and plan a project together.

Second, we are proud of being able to create a chatbot within the timespan. We were not too familiar with chatbot technologies but were able to nonetheless end up with a working chatbot. In addition, we were able to make a simple website in a day.

What we learned

We all learned a lot about Natural Language Processing and Chatbots. We learned about how to manage our project and use more advanced technologies.

What's next for Menthelp

Next, we would like to increase the usability of Menthelp. We would like to be able to save data, allows the user to create chats, and market our product further.