Ask HN: ML engineers of HN, how much do you code?

I'm a ML engineer myself. Most of the time I'm running experiments. I don't actually write a lot of code except when doing data processing. Other times I'm just fine-tuning BERT/GPT based models which does not require much code writing.

Is this the same situation for you? Any idea how we can bring more software engineering into this? I miss writing interesting/architecture code