Pre-Interview Cheatsheet
AI Engineer / Generative AI Engineer — Confidence Cheatsheet
A printable, focused refresher tuned for AI Engineer / Generative AI Engineer. Open the sections that matter to you and walk in confident.
Tuned for AI Engineer / Generative AI Engineer · Technology & AI > AI & Machine LearningRefresh Right Now The 60-second mental warm-up before you start.
- Know LLM basics, prompt engineering, RAG, embeddings, vector search, evaluation, safety and API integration.
- Understand hallucination, context windows, tokens, grounding, latency, cost and privacy.
- Refresh tool/function calling, guardrails, model selection and human-in-the-loop workflows.
- Strong AI engineers design reliable systems around models, not just clever prompts.
- Be ready to discuss how to evaluate an AI feature.
Core Vocabulary Terms interviewers expect you to use precisely.
- RAG: retrieval-augmented generation using external context.
- Embedding: numerical representation of text/objects for similarity search.
- Hallucination: plausible but false model output.
- Context window: amount of input/output a model can process.
- Guardrail: control to reduce unsafe or incorrect behavior.
Formulas & Frameworks The mental models that organise your answers.
- AI feature design: user task, data source, model, retrieval, prompt, evaluation, fallback, logging.
- RAG quality: chunking, metadata, retrieval, reranking, citation, answer evaluation.
- Evaluation: golden set, accuracy, groundedness, latency, cost, user feedback.
- Risk: privacy, security, bias, false confidence, overautomation.
Likely Interview Prompts Questions you should be ready for.
- How would you build a company document chatbot?
- What is RAG and why use it?
- How do you reduce hallucinations?
- How do you evaluate an LLM application?
- How do you control AI costs?
Red Flags To Avoid Common answers that lose interviews.
- Treating prompt engineering as the whole system.
- No evaluation plan.
- Sending sensitive data without controls.
- Ignoring failure modes.
- No fallback when retrieval fails.
What Sets You Apart Signals that move you from competent to memorable.
- Thinks in systems, data and evaluation.
- Can choose models based on task/cost/latency.
- Designs with grounding and auditability.
- Understands human review and safety.
30-Second Confidence Reset Anchor sentence to read just before you walk in.
AI engineering means turning probabilistic models into controlled products: ground the answer, evaluate outputs, manage cost and design safe fallbacks.