Technical practice
How would you frame a machine-learning problem for improving an inspection tool?
What to coverSuggestion: Start with the inspection outcome to improve and identify the data, labels and measurable success criteria needed. Distinguish a defect-classification task from a broader imaging-performance problem rather than assuming they require the same solution.
How would you analyze EBeam imaging and classification data before developing an algorithm?
What to coverSuggestion: Describe checks for data quality, class balance and patterns across wafer locations or defect types. Explain how those findings would shape the problem definition and an initial baseline.
How would you choose an approach for classifying wafer defects?
What to coverSuggestion: Compare a simple baseline with more complex machine-learning approaches in terms of available data, interpretability and performance. Explain what evidence would justify moving beyond the baseline.
How would you validate a prototype defect-classification concept?
What to coverSuggestion: Define a comparison against a baseline and use held-out data that reflects the intended inspection task. Discuss how you would examine errors, not just an aggregate score, before deciding whether the concept is promising.
What trade-offs matter when evaluating an inspection algorithm?
What to coverSuggestion: Discuss missed defects, false alarms and the usefulness of results for understanding manufacturing-process issues. Explain that the appropriate balance depends on the intended use and should be established during validation.
How would you investigate an algorithm that performs poorly on some inspection data?
What to coverSuggestion: Break down failures by data characteristics and check whether the issue comes from labels, the input data or the algorithm. Describe how a targeted experiment would test your leading explanation.
How would you research competing solutions to an inspection-data problem?
What to coverSuggestion: State the assumptions, expected advantages and limitations of each candidate approach. Propose a small, comparable test that would help select a concept for prototyping.
How would you turn a validated concept into software requirements?
What to coverSuggestion: Translate the intended inputs, outputs, performance measures and failure handling into requirements that can be tested. Distinguish what the prototype has demonstrated from what still needs verification in the inspection tool.
What changes when you productize a prototype algorithm?
What to coverSuggestion: Address reliable implementation, reproducible results and tests for expected and problematic inputs. Explain how you would check that the programmed solution preserves the behavior observed during validation.
Behavioural practice
Tell me about a time you had to define an unclear technical problem.
What to coverSuggestion: Use a real situation, the questions and analysis you used to narrow its scope, and the result. Focus on how you made the problem actionable rather than claiming certainty too early.
Tell me about a time evidence changed your proposed solution.
What to coverSuggestion: Describe the original concept, the test or research that challenged it, and the action you took. Give the result and what the evidence taught you about evaluating alternatives.
Tell me about a time you explained a technical concept so others could act on it.
What to coverSuggestion: Set out the audience and the decision or requirement they needed to understand. Explain how you communicated trade-offs, what action followed and the result.
Tell me about a time you moved a prototype or analysis toward a dependable implementation.
What to coverSuggestion: Give the situation, the validation or implementation gap you found, and the steps you took to address it. State the outcome, including any limitations that remained.