API-based deployment in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding API-based deployment benefits

Which of the following is the primary advantage of deploying a machine learning model via an API?

A It eliminates the need for any server or cloud infrastructure.

B It automatically improves the model's accuracy over time without retraining.

C It allows real-time access to the model from different applications without sharing the model code.

D It stores the entire training dataset within the API for faster predictions.

💡 Hint

Think about how APIs let different programs talk to each other.

APIs provide a way for applications to send data to the model and get predictions without exposing the model's internal code or requiring local installation.

❓ Predict Output

intermediate

Output of a simple API call to a deployed model

Given the following Python code snippet calling a deployed ML model API, what is the printed output?

Prompt Engineering / GenAI

import requests
response = requests.post('https://api.example.com/predict', json={'input': [1, 2, 3]})
print(response.json())

A {"prediction": [0.1, 0.5, 0.4]}

B {'error': 'Invalid input format'}

C {'prediction': [1, 0, 1]}

D requests.exceptions.ConnectionError

💡 Hint

Assume the API is working correctly and returns prediction probabilities.

The API returns a JSON with prediction probabilities for the input data, so the output is a dictionary with a 'prediction' key and a list of floats.

❓ Hyperparameter

advanced

Choosing timeout settings for API deployment

When deploying a machine learning model via an API, which timeout setting is most important to ensure good user experience without overloading the server?

A Set timeout to zero to prioritize requests in the order they arrive.

B Set a very high timeout so the server never cancels requests, even if slow.

C Disable timeout settings to allow unlimited request processing time.

D Set a low timeout to quickly reject requests that take too long, preventing server overload.

💡 Hint

Think about balancing responsiveness and server health.

A low timeout helps reject slow requests early, keeping the server responsive and preventing overload from stuck or slow calls.

❓ Metrics

advanced

Evaluating API latency metrics

You deployed a model via API and collected these latency times (in milliseconds) for 5 requests: [120, 150, 130, 200, 170]. What is the average latency?

A 154 ms

B 170 ms

C 150 ms

D 1540 ms

💡 Hint

Calculate the sum of all latencies and divide by the number of requests.

Sum is 120+150+130+200+170 = 770 ms. Divide by 5 requests = 154 ms average latency.

🔧 Debug

expert

Identifying cause of API deployment failure

After deploying your ML model as an API, clients report a 500 Internal Server Error when sending valid requests. Which of the following is the most likely cause?

A The client is sending requests with wrong input data types.

B The model file path is incorrect or missing on the server.

C The API endpoint URL is misspelled in the client code.

D The server has no internet connection.

💡 Hint

500 errors usually mean server-side problems, not client mistakes.

A 500 error indicates a server error, often caused by missing model files or code errors during prediction, not client input or URL issues.