In , ML interviews focused on classical algorithms. You'd explain gradient descent, implement logistic regression, and discuss bias-variance tradeoff. LLMs were niche.
By , the field shifted. Interviewers now expect you to explain transformer attention, discuss hallucination mitigation, and design RAG systems. Classical ML still matters, but it's now foundation rather than focus.
The shift in numbers:
- LLM/GenAI questions: % → %+ of technical rounds
- System design: Includes ML serving and inference optimization
- Behavioral: Now probes AI safety and ethics positions