You just got hired at a new bank. You walk into your first day on the job and the CEO hands you a bank statement. He tells you to go through the financials and write a report on whether we should give them a business loan. You ask, “how do I know if he is a good fit or not?” and he looks at you like you’re crazy. “Obviously by telling me the total monthly deposits and withdrawals. Maybe give me some insights on their big regular bills like rent, employee salary, utilities, ect. Oh and try to catch if they have any other outstanding loans! But overall just use common sense on if you’d give them a loan or not based on how they spend their money.”
You think to yourself — Okay that makes some sense. So you walk over to talk to your first coworker. He’s frantically writing notes and highlighting sections on a stack of papers. Thats when you realize … he’s manually going through every single transaction on the bank statement! And he has another 20 packets to get through before his day is over! Then you scan the room and realize there are 30 of your co-workers reading statement after statement doing the exact same thing!
You know there has to be a better way. Being the intelligent, well-trained Machine Learning Engineer you are, you start thinking of ways to gather data, so you ask your co-worker if you might be able to borrow some of his bank statements. He side-eyes you — you’re the new guy trying to take all the credit in his mind.
… Okay, guess you’ll have to use a different approach then . . .
The Challenge
Your challenge is to design an ML-based system that is able to analyze and provide insights given a bank statement. In this challenge, we have provided you with 4 bank statements (the one given to you by the CEO + 3 more he decided to give to you later) but the type of bank statement used to evaluate your system can come from anywhere. This problem is meant to simulate the real world scenarios you would face working @ a new startup with new people. You will be scored based on the creativity and ingenuity in delivering an MVP to this problem. This challenge will not only test you on your ability to use Artificial Intelligence, but also test you on the way you present data via a user experience, the way you source data in restrictive environments, and the way you scope problems to account for both edge-cases in the solution and the timeline of the challenge.
There are no rules or restrictions on the technologies you can use or the medium in which you deliver your solution. You can scope this problem however you see fit as long as fundamentally the solution is compelling. Be creative.
Remember, this is an MVP. So lean into your strengths as a developer whatever they may be. If you can prove to the CEO that this is a worth while project then its likely he’ll help you scale your solution and give you a team that has your back (like us @ Cascading AI 😄)