IBM A1000-125 - Assessment: AI Engineer Intermediate Practice Exam: Medium Difficulty 2025

A data scientist is building a classification model and notices that the training accuracy is 98% while the validation accuracy is 72%. The model performs poorly on new data. Which combination of techniques would be most effective to address this issue?

An AI engineer needs to implement a Watson Assistant chatbot that handles customer inquiries and seamlessly transfers complex cases to human agents. The chatbot must also access a product database to provide real-time inventory information. Which Watson Assistant features should be combined to achieve this requirement?

A machine learning model deployed in production is experiencing concept drift, where the statistical properties of the target variable have changed over time. What is the most appropriate strategy to maintain model performance?

An organization is using Watson Discovery to analyze thousands of technical documents. Users report that search results often miss relevant documents that use different terminology. Which Watson Discovery feature should be configured to improve semantic search capabilities?

During model training, a data scientist observes that the loss function decreases steadily for the first 20 epochs, then oscillates wildly without converging. The model uses stochastic gradient descent. What is the most likely cause and solution?

An AI engineer needs to choose an appropriate evaluation metric for a fraud detection model where fraudulent transactions represent only 0.5% of all transactions. The business priority is to catch as many fraud cases as possible while keeping false alarms manageable. Which metric should be prioritized?

A company wants to deploy multiple versions of a machine learning model simultaneously, routing 90% of traffic to the stable version and 10% to a new experimental version to compare performance. Which deployment strategy should be implemented?

An AI engineer is implementing Watson Natural Language Understanding to extract sentiment and entities from customer reviews. The default model performs well on general product reviews but poorly on domain-specific technical products with specialized terminology. What approach should be taken?

A neural network model for image classification exhibits high bias, with both training and validation accuracy plateauing at 65% when the target is 90%. Which approach would most effectively address this underfitting problem?

A production machine learning pipeline processes real-time data streams and makes predictions. The operations team needs to monitor model health and detect when predictions become unreliable. Which monitoring approach should be implemented?