Name the project : „#more“ Elevator pitch: Transactions is often cryptic data to the user and the user is often cryptic data to the bank. We want to add value adding information to transactions in form of tags. Tag are generated by a neuronal network algorithm or the user input and used for the bank to get a tag-footprint of the user and for the user to keep track of expenses and trends.
Type of project: Crossplattform web and mobile app, intelligent tagging algorithm MVP: self-learning cloud-based neural network algorithm as service for existing bank apps IoS/ Android app with tagged transaction overview Target group: Banks to generate valuable customer data -> know your customer Technology affine users who are looking for a nice opportunity to keep track of expenses (business expenses, tax return) Use Case: Banks: Can request a user up-to-date tag-footprint before a interview User: mobile application
Challenges & solutions: The tagging algorithm is aself-learning neural network witch is not easy to setup. A huge set of data is needed to teach the network to be really good and intelligent. We generated a lot of sample data and tried many neural network architectures to find a suitable one. User inputs needs to be easy – used Gini API , so the user just needs to take a photo of the bill to tag the transactions
Scalability: webapp is easily scalable and server capacity will scale with number of users tagging algorithm needs strong server architectures, but will not scale 1:1 with user (teaching is not scaling a lot with user count)
Link&Code : https://github.com/daliwali/burda
Team: Sven Eliasson (Algorithm development & product engineer), Sven.Eliasson@gmx.de Evi Soeyler, (bussniess development) Dali Zheng, (backend engineer) email@example.com Vadym Yatsyuk, (web and mobile devceloper) Vasp3D@gmail.com