Technical practice
How would you architect a generative AI solution from data pipeline through production monitoring?
What to coverSuggested preparation: Explain how you would connect data preparation, model or application development, evaluation, deployment and monitoring. Discuss design choices that would change with scale, performance needs or the use case.
How would you design and evaluate a retrieval-augmented generation application?
What to coverSuggested preparation: Cover document preparation, retrieval, use of a vector database and how retrieved content reaches the LLM. Explain how you would assess both retrieval quality and the resulting answers; FAISS, Pinecone and Weaviate are listed as nice-to-have familiarity, not required tools.
When would you use prompt optimization rather than fine-tuning or model adaptation?
What to coverSuggested preparation: Compare the data, effort, evaluation and operational implications of each approach. Describe how you would test whether an adaptation improves the intended task enough to justify its complexity.
How would you make an AI-agent workflow dependable in production?
What to coverSuggested preparation: Explain how you would structure orchestration, constrain agent actions and handle errors or uncertain outputs. Address how you would evaluate the workflow as a whole, not just its underlying model.
How would you approach deploying an AI solution with AWS Bedrock and SageMaker?
What to coverSuggested preparation: Describe which parts of an example solution you would place in each service and why, without assuming the employer uses a particular architecture. Cover deployment, evaluation and operational considerations for a production-grade system.
How would you build and validate a forecasting or prediction model?
What to coverSuggested preparation: Discuss an appropriate statistical, machine-learning or deep-learning baseline, relevant features and a validation scheme that avoids leakage. Explain how you would compare model quality with complexity and operational cost.
How would you prepare large-scale data for ML using SQL and Python?
What to coverSuggested preparation: Explain how you would check data quality, construct reproducible features and manage processing at scale. Discuss when distributed processing might help; Spark is listed as nice-to-have exposure.
How would you evaluate an LLM-powered application before release?
What to coverSuggested preparation: Define measures tied to the application’s task, alongside checks for reliability and problematic outputs. Explain how you would use representative test cases and interpret trade-offs rather than relying on a single score.
How would you troubleshoot a production AI system that becomes slower or less effective?
What to coverSuggested preparation: Describe what you would monitor across data, model, application and infrastructure, then how you would isolate the cause. Discuss safe changes, versioning and CI/CD as possible operational practices; MLOps experience is listed as nice to have.
How would you incorporate responsible AI into a system design?
What to coverSuggested preparation: Identify risks appropriate to the use case and explain how you would evaluate and mitigate them throughout development and deployment. Distinguish practical checks from formal governance frameworks, which the posting lists as nice-to-have understanding.
Behavioural practice
Tell me about a time you mentored an AI engineer or data scientist.
What to coverSuggested preparation: Use a real situation, your specific coaching actions and the result. Explain how you supported the person’s work while maintaining technical standards.
Describe a time you worked with product or business colleagues to define an AI solution.
What to coverSuggested preparation: Give the situation, how you clarified the problem and constraints, and the result. Show how you communicated technical trade-offs without assuming a particular company process.
Tell me about a time you led an AI project from design toward production.
What to coverSuggested preparation: State the situation, your leadership decisions and the outcome, including what you personally owned. Explain how you coordinated work across development, evaluation and deployment.
Describe a production AI setback you had to address.
What to coverSuggested preparation: Use a real situation, the steps you took to investigate and respond, and the result. Highlight how you prioritized reliability, performance or model quality under the circumstances.
Tell me about a time you raised a responsible-AI concern or challenged a proposed approach.
What to coverSuggested preparation: Describe the situation, the evidence you used and how you worked with others to decide what to do. State the outcome without overstating your authority or the impact.