This case is about testing candidates applying for a Solution Engineer position (data science track) at Mindee. The goal is to deliver high performances text recognition models trained using our open-source OCR library docTR.
Read this article to understand the use case.
Ask the hiring manager to give you the training data. If you don't have enough computation capability, try to train on a smaller subsample of data.
Show us how your model is performing during a live demo.
What would you change / add in your training process to optimize the results?
During your presentation, we will:
- review your code quality using your training and inference scripts
- check your ability to deliver a end-to-end demo
- ask you questions about the models and how to optimize it
Good luck!