Phenom Interview Question

11. Q: Design a neural network for multi-class classification using text, categorical, and numerical inputs. 12. Q: Why is ReLU preferred over sigmoid in hidden layers? Explain the vanishing gradient problem. 13. Q: Explain Random Forest. Why is it called "Random," and how does it reduce overfitting? 14. Q: What is Feast? Why would you use a feature store in an MLOps platform? 15. Q: Explain your MLOps platform architecture, including model versioning, pipeline orchestration, deployment, and monitoring. 16. Q: Write code to find the row with the minimum sum in a matrix without using Python built-in functions. Explain the time complexity. 17. Q: How do you prioritize accuracy, latency, infrastructure cost, and customer requirements when building an AI product? 18. Q: What is planning in an AI agent? Why is planning important, and how does it differ from a simple workflow? 19. Q: If a tool invoked by an agent fails, how should the agent recover and continue execution? 20. Q: Explain the complete workflow of your NL-to-query (Spotter) agent, including retrieval, validation, query generation, execution, and feedback.