Inspiration

We considered the annoyance of waiting for a reimbursement or refund. Thanks to Attest, we were able to run a market research survey consisting of 4 questions on 250 people. We were to able to validate the reimbursement problem before attempting to solve it.

What it does

We reimburse employees instantly with computer vision (up to a certain limit) and human verification (for larger sums) by integrating bank accounts (using ThoughtMachine for the demo). This means they don't have to wait around 2 weeks to 2 months for reimbursements. Companies would use us to reduce the pain points with calculating expenses for their employees, as well as the burden on administration and management. For the end user, they get their money instantly. What more could they want from a reimbursement?

How we built it

We prototyped the computer vision (Google vision, and some in house modifications) and banking code (ThoughtMachine) in Jupyter notebook, then built a Flask backend hosted locally, exposed by ngrok. Then it was deployed to GCP/ firebase. We built the mobile app with react native, including react navigation for seamless navigation, lottie for animations and react-native-firebase, for authentication with firebase. We also built a dashboard for administrators to approve unusual or expensive reimbursements. Lots of this happened in parallel or through pair-programming, speeding up problem solving and development. We had regular stand ups (every 1-3hrs), and recommend the same.

Challenges we ran into

Using expo in react native was a very big problem. We had to revert to using the bare react-native workflow. We found another team who used expo, and helped them with their problems at 4am in the morning. We also had trouble using google cloud storage buckets from a python backend to serve images.

Accomplishments that we are proud of

Building the a working app interfacing with a backend. Setting up a nice GoogleVision/GCS/Firebase stack as newcomers to these technologies

What we learned

More skills in react native, python, web requests (especially images and form-data). More skills on Google cloud products and discovery of the Vault tool from Thought Machine

What's next for Reimbursr

Integration with bank accounts. Improve receipt scanning algorithms. More controls for backoffice.

Share this project:

Updates