What fundamentals should AI engineers brush up on before interviews?

Most candidates focus on tools. Strong candidates focus on fundamentals.

1) LLM basics (non-negotiable)

Context limits, tokenization, cost/latency, and failure modes (hallucination, prompt sensitivity)

2) Retrieval (RAG is table stakes)

Embeddings, vector search tradeoffs, chunking, grounding

3) System design thinking

Pipelines: Input → Retrieval → Generation → Verification

Plus caching, cost, and reliability

4) Evaluation (biggest gap)

Not just accuracy → faithfulness, usefulness, safety

Bottom line

If you don’t understand how the system fails and how to measure it, you’re not ready.