Career advice: Transitioning from traditional ML to LLM engineering

I've been a traditional ML engineer for 6 years (scikit-learn, PyTorch, classical NLP) and want to transition to LLM-focused roles. The job market seems to have shifted dramatically.

Questions for those who've made the transition:

1. How much of traditional ML knowledge transfers? 2. What skills are most valued? (RAG, fine-tuning, prompt engineering?) 3. Are companies looking for PhD-level researchers or practical builders? 4. What projects would you recommend building for a portfolio?

I've been working through the OpenAI cookbook and building RAG systems, but not sure if that's enough to be competitive.