I'm new to the tech world and recently joined an AI software startup as a backend intern. However, I'm currently the only backend developer here, and I'm facing a tough situation. The entire backend was created using AI before I came onboard, and I'm finding it quite a challenge to work with. There's no documentation, very little logging, and I'm unclear about the functionality of many APIs, as well as the structure of the database. My tasks include understanding and organizing the existing backend code, cleaning up the database, adding proper documentation, and gradually fixing and refactoring the code where necessary. I'm really not sure where to begin and could use some experienced guidance on how to tackle this chaotic backend and database effectively.
3 Answers
You're in a tough spot as an intern, and it sounds like your workload is heavy. Try to manage the cognitive load by breaking down the tasks. Start by writing regression tests to ensure existing functionality remains intact after you make changes. Focus on understanding the code in chunks—break down the modules and figure out what can be improved first. You might notice some inefficiencies, like N+1 queries or missing caches, which you can list out for quick fixes as you go along. If the code's really convoluted, don't hesitate to rewrite complex components for clarity—maybe use declarative functions to keep it organized.
It sounds like your company dropped the ball by leaving you to handle this AI-created backend alone. It's tough to be in this position, but you're going to have to navigate through this mess yourself.
Have you thought about asking the AI that built the backend to generate some documentation for you? It should be able to handle that without a hitch!
I wouldn't typically recommend relying on AI for this, but honestly, it's pretty unreasonable to expect you to deal with such a messy situation alone. If you can use AI to help document the code, go for it!
This situation really does feel like a sink or swim challenge! But the regression tests seems like a solid starting point. By interacting with each endpoint, you'll get a clearer picture of how things work before diving into changes. Good luck; AI-generated code can be a mixed bag!