I had exactly the same problem and never quite solved it despite ending up as a research scientist at a world-class university. My best advice is to find an en

I had exactly the same problem and never quite solved it despite ending up as a research scientist at a world-class university. My best advice is to find an engineering-oriented advisor, even if they are from another department. If that fails, find an interesting engineering problem, say "Execution of a distributed rule set across a distributed knowledge base". What are the fundamental problems? In this case, they are speed and consistency. Researchers don't care about speed, but they do care about consistency. So your topic becomes "Maintaining consistency in a dynamic distributed rule network". You solve this one theoretical problem, engineer a cool system to illustrate it, and get a PhD. Best wishes and good luck.