Mood Mate

Many people face mental health challenges on a daily basis but either don't have time or don't have access to therapy. Mood Mate attempts to bridge this gap by providing AI analysis for journaling.

Comment

Inspiration

Since starting work at an internship this summer, I have not had time during the day to attend therapy. Many people face the same issue, or don't have the option to go to therapy. I built Mood Mate to provide a solution to this problem. While it is not a one-to-one replacement for therapy, it is capable of providing insights and suggestions based on a user's journal entries.

What it does

Mood Mate provides an intuitive and clean user interface that allows users to create journal entries. These entries are then fed to a bert text classification model which extracts tags based on the input. Then, the entries and those tags are sent to Google's Gemini API. This is where the insights are generated and provided to the user.

How we built it

I used HP AI Studio to modify and publish the bert model. I also used it over the course of development to test various aspects, like my API endpoints. ML Flow and its direct integration with my Jupyter notebook allowed me to easily push changes to my model and use the API to send and receive data in my webapp. The frontend was built with Vercel's v0.

Known Issues / Limitations

  • MLflow model serving is not connected live to the frontend. While the backend works locally, it has not been deployed persistently due to time constraints and system setup issues within AI Studio
  • Frontend prediction calls will not return results unless the model is manually served using mlflow models serve.

What Works

  • BERT model training, logging, and predictions via MLflow locally
  • Frontend UI for text input and emotion analysis layout
  • FastAPI or MLflow backend works in isolation

Accomplishments that we're proud of

Overall, I am proud that I was able to get this spun up. This is the first project I have made myself using machine learning, so it was a nice head-first dive into the field!

What we learned

I learned a lot about model development and publishing, and the training process.

What's next for Mood Mate

I plan to keep developing this as I feel there is a need for this type of application in our fast-paces world. One feature I wanted to add but did not have time for was paper journal uploads. I will be implementing that in the future for sure!

Built With

Share this project:

Updates