> research isnt something that happens at startups

Entirely depends on the startup and the product. R&D happens on many startups. Unless someone has a limited exposure on AI and ML-oriented startups, this is far from truth. OpenAI is an applied research company. They produce research papers and puts it into production. In the electronics department, OnePlus has risen as a great R&D startup capable of producing rapid R&D-based products. Grammarly puts a ton of money on its R&D to create a more domain-specific GPT model because it is vital to their product.

> The divide you describe

One does not need to probe deeper into this. Ask an experienced Data Engineer, a Data Scientist, and a DevOps. There is a clear DISTINCTION of what they do and how they balance each other. The divide isnt hostile. It's more of "we want this, you cant have all of this type of relationship, besides the usual difference of who works with what.