Luciano Krebs

I contributed to a multidisciplinary team to deliver multiple versions of an API responsible for approving or not a given credit request.

We used Python, ML, Redis, FastAPI, and ECS to run the app in production. As we were iterating over the current model and using more data to improve the numbers, we could keep the credit default under control and make the company profitable.

I owned the publication of a React Native app that helps educational companies manage their businesses to the App Store and PlayStore.

This app's main functionalities included:
• Camera to upload personal documents
• Push notifications to update the users about their credit requests
• Location for fraud prevention
• Contacts to make the sharing experience easy
• Social authentication using Google, Facebook, GitHub, and email and password

The app was our first public release and was responsible for our initial traction. Using it, we made our first ten million in loans, validated the idea, and raised a VC seed round two months later. Months later, we pivoted our solution to be web-only instead of mobile-only and ended up removing the app from the stores.

As we needed to validate our idea and product directed to educational businesses, our initial back end was a monolithic Node.js and Express app. Once we found the product-market fit and started to scale, we split the primary server into eight microservices.

After the transition finished, we achieved the following results:
• Decreased the number of new bugs by around 10%, making the ratio between bugs fixed vs. new bugs greater than one for the first time
• The release cycle went from once a week to daily releases as the team became more autonomous
• Increased the overall test coverage by around 20%, as we wrote more tests during the microservices development