Technical practice
How would you turn a customer's long-term AI vision into an engineering roadmap?
What to coverSuggested preparation: explain how you would learn the goals of executives, operators, and technical teams, then identify dependencies and near-term deliverables. Distinguish a strategic aspiration from an application that can be shipped and measured.
How would you prioritize AI applications for a customer?
What to coverSuggested preparation: compare potential business value, user pain, data readiness, implementation effort, and operational risk. Explain what evidence would justify building one application first and what outcome would show it worked.
How would you design a production agent that uses tools across multiple steps?
What to coverSuggested preparation: describe how the agent would choose tools, manage intermediate results, and handle failed or unexpected tool responses. Discuss when a constrained workflow may be preferable to a more autonomous agent; tool use and multi-step agents are listed as bonus experience.
When would you use a RAG pipeline, and how would you evaluate it?
What to coverSuggested preparation: explain when retrieving customer information is useful, how you would assess retrieval quality and answer grounding, and how you would handle missing or conflicting information. RAG is listed as bonus experience, not a stated requirement for every project.
Where would you place human review in an AI workflow?
What to coverSuggested preparation: identify decisions that warrant approval or escalation and explain the trade-off between oversight and workflow speed. Describe how users would see enough context to make a decision; human-in-the-loop workflows are listed as bonus experience.
How would you judge whether a shipped AI workflow succeeds?
What to coverSuggested preparation: separate technical measures, such as task quality and failure rates, from adoption and measurable customer outcomes. Explain how you would test with real users and data rather than treating a working prototype as proof of production value.
How would you investigate an agent that performs well in tests but fails for end users?
What to coverSuggested preparation: walk through reproducing failures and checking inputs, retrieval, tool responses, and workflow steps. Explain how findings from operators or end users would guide a fix and how you would check for regressions.
What would you consider before deploying an LLM-powered workflow to the cloud?
What to coverSuggested preparation: discuss reliability, access to customer data, monitoring, and recovery when a dependency fails, alongside the pressure to ship quickly. Cloud deployment and DevOps tooling are bonus qualifications; no particular provider or deployment method is required.
How would you turn a customer-specific solution into a reusable Mosaic capability?
What to coverSuggested preparation: identify which features, patterns, or abstractions could serve other engagements without obscuring customer-specific needs. Explain how you would decide whether a repeated need merits a toolkit change rather than another one-off implementation.
How would you keep implementation quality high while shipping quickly?
What to coverSuggested preparation: describe which tests, reviews, and operational checks you would retain under time pressure and where you could simplify scope instead. Python or TypeScript foundations are listed as a bonus, so focus on sound engineering choices rather than assuming a mandated language.
Behavioural practice
Tell me about a time you took ownership of an ambiguous problem without waiting for an assignment.
What to coverSuggested practice: give the situation, the next step you chose, and how you communicated that decision to others. State the result and what you changed when new information emerged.
Describe a time you aligned people with different goals around a technical direction.
What to coverSuggested practice: identify the stakeholders and their competing concerns, then explain the actions you took to reach a decision. Give the result, including how you kept people informed rather than working in a silo.
Tell me about a time direct feedback from users changed what you built.
What to coverSuggested practice: describe the user's pain point, how you learned about it, and the change you made. Explain the observed result without substituting shipped features for user outcomes.
Describe a time you had to balance delivery speed against quality.
What to coverSuggested practice: set out the deadline or uncertainty, the risk you weighed, and where you chose to move quickly or slow down. Give the result and explain what made the trade-off appropriate.
Tell me about a time you gave candid feedback while sustaining a collaborative relationship.
What to coverSuggested practice: describe the situation, the specific feedback or decision you communicated, and how you treated the other person. Explain the result and what the example shows about working with urgency and kindness.