Interviews
Common AI & machine learning interview questions
The concepts ML engineers and applied scientists are most often asked about — and how to answer well.
8 min read
AI and ML interviews blend theory, coding and product judgment. Expect questions in these areas:
Fundamentals
- Bias–variance trade-off: explain under- vs. over-fitting and how regularization, more data or simpler models help.
- Evaluation: when to use precision/recall, ROC-AUC, PR-AUC, F1, calibration and business metrics.
- Data leakage: how it happens (target leakage, time leakage) and how to prevent it.
Deep learning & LLMs
- How attention works and why transformers scale.
- Fine-tuning vs. retrieval-augmented generation (RAG) vs. prompting — when to use each.
- Evaluating LLM outputs: offline eval sets, human review, and guardrails for safety and hallucination.
- Serving concerns: latency, batching, quantization and cost per request.
ML system design
You may be asked to design a recommendation system, a fraud detector or a search ranker. Cover: problem framing, labels, features, model choice, offline and online evaluation, deployment, monitoring for drift, and feedback loops.
Tips
- Tie every technical choice back to the product goal.
- Be honest about what you haven't used — then explain how you'd learn it.