According to the 2017 survey, CareerBuilder, a leading job site, found some startling statistics related to debt, budgeting and making ends meet.
- 78 percent of U.S. workers live paycheck to paycheck to make ends meet
- Nearly one in 10 workers making $100,000+ live paycheck to paycheck
- More than 1 in 4 workers do not set aside any savings each month
- Nearly 3 in 4 workers say they are in debt today - more than half think they will always be
- More than half of minimum wage workers say they have to work more than one job to make ends meet
We wanted to build an Ai financial assistant that allows individuals to save effortlessly and reach their monthly saving milestones to pay off debt faster and get on the path to financial freedom.
What it does
SaveWise is an AI service that connects to a user's bank account using the Plaid API, trains on their past transaction data and predicts upcoming expenses and savings potential for the future. Users who reach their monthly saving goals have the ability to pay off debt faster incurring lower interest and other fees over time.
SaveWise also delivers proactive notifications based on its forecasts allowing the user to optimize their spending patterns and save money on a daily basis. It also suggests more economical alternatives for meals, transportation, entertainment and shopping requirements using Yelp local search API.
How we built it
The service was built using a bunch of technologies and frameworks:
- Account, Transaction and liability data is collected using the Plaid API
- The Voice AI is powered by Houndify
- The prediction models are trained using RNNs with Pytorch
- Local deals search is powered by Yelp API
- Backend is built in GO and Frontend is built using Angular
Challenges we ran into
As we took an ambitious goal to build a category wise spending prediction AI service, and consolidate all this intelligence into a platform-agnostic web application for anyone to use, we ran into a few challenges:
- Integrating all the microservices into an application and formatting data for ease of use
- Training prediction model for each category with a limited dataset.
- Creating accurate personalized prediction.
- Serving the backend with all the microservices on AWS
Accomplishments that we're proud of
Building a relevant solution that can be impactful for over 230 million Americans allowing them to achieve financial wellbeing is an inspiring pursuit. With this bold ambition in mind, we were to able to build and deploy an end to end functional application leveraging the tools and services provided by our sponsors like Plaid. We were able to build a lot of functionality with a very short time frame. We were able to onboard the plaid API, integrate Houndify voice AI, recommend daily savings and train our models using Pytorch.
What we learned
The best way to build an AI service is to use multiple microservices to get started and identify the core use cases for end-users that can deliver maximum impact.
What's next for SaveWise
The following features are in our development pipeline:
- Add more intelligence around the user's surroundings to recommend the most economical alternatives for all requirements.
- Optimize the models to predict and suggest maximum monthly savings.
- Allow user to update notification settings so that he/she is notified an unwise purchase/transaction.
- Suggest investment options based on savings accumulated.