Inspiration
We were interested in doing something financially related using machine learning and natural language processing
What it does
We created a Google Action for Google Assistant so that a user can record his/her purchases and Google Assistant would keep track of his/her monthly budget. We also created a custom ML model to determine whether it would be feasible to invest in stocks given the current budget and money spent. In addition, our platform integrates ethereum blockchain technology to easily transfer money between two users. Our last feature is the use of Dialogflow's Voice service. In addition to talking to Google Assistant through Google Home, users can also dial a number to access our service from anywhere with cellular service.
How we built it
We used Dialogflow to configure the different intents and routes, using two Node.JS servers to properly handle the requests. We relied heavily on firebase and firebase cloud functions to easily execute functions. Our machine learning model is run on a separate server and was created using Tensorflow.
Challenges we ran into
We really wanted to train and host our ML model using Google's ML engine, but we were not able to configure our model to fit the requirements of the cloud service. We ended up training our model locally and building a custom server to host it. In addition, Dialogflow's integration with Firebase Cloud Functions was very hard to understand and hard to debug.
Accomplishments that we're proud of
We are proud of actually getting the phone aspect of our project to work
What we learned
To construct models with the intent of training them on the cloud instead of however we want to
What's next for personalBanker
We would like to make the GUI more beautiful and easy to use
Built With
- css
- dialogflow
- express.js
- firebase
- javascript
- node.js
- pug.
- python
Log in or sign up for Devpost to join the conversation.