Personal assistant agent patterns in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the role of memory in personal assistant agents

Which of the following best describes why memory modules are important in personal assistant agents?

A They allow the agent to remember past interactions to provide context-aware responses.

B They speed up the agent's processing by caching all user data permanently.

C They replace the need for natural language understanding in the agent.

D They ensure the agent never forgets any information, even irrelevant details.

💡 Hint

Think about how remembering past conversations helps a friend assist you better.

Memory modules help personal assistant agents keep track of previous user inputs and context, enabling more relevant and personalized responses.

❓ Model Choice

intermediate

Choosing the right model architecture for task management

You want your personal assistant agent to manage calendar events and reminders effectively. Which model architecture is best suited for this task?

A A sequence-to-sequence model with attention to handle natural language commands and generate structured calendar entries.

B A convolutional neural network trained on image data to recognize calendar layouts.

C A reinforcement learning model that learns to play chess.

D A simple linear regression model predicting numerical values.

💡 Hint

Consider models good at understanding and generating sequences of text.

Sequence-to-sequence models with attention are effective for converting user commands into structured data like calendar events.

❓ Metrics

advanced

Evaluating personal assistant agent response quality

Which metric is most appropriate to evaluate how well a personal assistant agent understands and responds to user queries?

A Confusion matrix for spam detection.

B BLEU score measuring similarity between generated and reference responses.

C Accuracy of image classification on a test set.

D Mean squared error between predicted and actual numerical values.

💡 Hint

Think about comparing text outputs to expected answers.

BLEU score is commonly used to evaluate the quality of generated text by comparing it to reference responses, suitable for personal assistant outputs.

🔧 Debug

advanced

Debugging a multi-turn dialogue failure in a personal assistant

Given the following simplified dialogue state update code snippet, what is the main issue causing the agent to lose track of user context?

def update_state(state, user_input):
    if 'reset' in user_input:
        state = {}
    else:
        state['last_input'] = user_input
    return state

state = {'last_input': 'hello'}
state = update_state(state, 'what is my schedule?')
state = update_state(state, 'reset')
state = update_state(state, 'do I have meetings?')
print(state)

A The state dictionary is mutated correctly, so no issue exists.

B The function does not handle the 'reset' command properly, so state is never cleared.

C The state variable is reassigned inside the function but not updated outside, causing loss of context.

D The function returns None instead of the updated state.

💡 Hint

Consider how Python handles variable assignment inside functions.

Reassigning the state variable inside the function creates a new local dictionary, but the caller's state remains unchanged, causing context loss.

❓ Hyperparameter

expert

Optimizing reinforcement learning for personal assistant task scheduling

You are training a reinforcement learning agent to schedule tasks for users. Which hyperparameter adjustment is most likely to improve the agent's ability to balance immediate and future rewards?

A Reducing the number of training episodes to avoid overfitting.

B Decreasing the learning rate to zero to stop learning.

C Setting the batch size to 1 to update weights after every single step.

D Increasing the discount factor (gamma) closer to 1 to value future rewards more.

💡 Hint

Think about how the agent values rewards over time.

The discount factor controls how much future rewards influence decisions. Increasing it helps the agent consider long-term benefits.