Data & Machine Learning
ML System Design & MLOps
8 practice questions. Free questions open a full answer guide; the rest unlock with Pro.
- Once a model is deployed, how do you know when it's degrading, and what's the difference between data drift and concept drift?
- Your model scores well in offline evaluation but performs noticeably worse once it's serving live traffic. What would you investigate, and how do you prevent this class of problem?Go Pro
- Your model scores well in offline evaluation but the team keeps finding that a feature computed in the training pipeline doesn't match the value the serving path produces for the same entity. How do you design the system so training and serving features stay consistent?Go Pro
- How do you make a model training run reproducible, so a teammate (or you, three months later) can recreate the exact same model?Go Pro
- You're standing up a model-serving platform and need a safe way to ship new model versions, because the team has been pushing retrained models straight to production and occasionally something gets quietly worse. How do you design the rollout and rollback so a bad model can't take down quality?Go Pro
- You own a model that's served live in production. How do you decide when it has degraded enough to retrain, and how do you distinguish a shift in the input data from a genuine drop in the model's predictive quality?Go Pro
- What is training-serving skew in an ML system, and how do you keep the features used in training consistent with the features used at inference?Go Pro
- How would you roll out a new version of a model that's already serving production traffic so that a regression can't quietly hurt users?Go Pro
Want questions matched to your role? Paste a job title, job description, or CV for a personalized set, or go Pro to unlock the full bank.