Why is machine learning 'hard'? by/via @zaydenam

Machine learning engineers command higher salaries than software engineers, probably because at its core, machine learning is a fundamentally harder debugging problem than standard software. It's not the math; it's that machine learning adds two additional (potentially bug-infested) dimensions: the model and the data.

Great article by Zayd which I think needs to be shared.  I think the key to fixing issues and getting ML working is you need:
1. Focused Concentration.
2. Knowledge of ML, DL,
3. A good level of understanding of the input data and it's model
4. Experience (sometimes you have come across something like this before so just know it is likely to be the same thing).

Don't be discouraged - the more you do with ML the better you will be at debugging and fixing errors with any code you have developed.