AI Engineer Interview Questions

A comprehensive guide covering Core ML, Deep Learning, LLMs, NLP, System Design, MLOps, and Computer Vision for AI and Machine Learning roles.

Total Questions:150
Difficulty Levels:
BeginnerIntermediateAdvanced
0%

Overall Progress

0/150

1.Explain gradient descent like I'm 5 years old.

2.What's the difference between L1 and L2 regularization and when do you use each?

3.Walk me through how backpropagation works.

4.What is overfitting and how do you prevent it in real projects?

5.Explain precision vs recall - when do you optimize for which?

6.What's the bias-variance tradeoff?

7.How do you handle imbalanced datasets?

8.What's cross-validation and why do we use it?

9.Explain the difference between bagging and boosting.

10.When would you use a decision tree vs neural network?

11.Explain how a neural network learns.

12.What activation functions do you know and when do you use each?

13.Why is ReLU better than sigmoid?

14.What's batch normalization and why use it?

15.What's dropout and how does it work?

16.Explain CNN architecture - what are convolution and pooling layers?

17.What's the difference between RNN, LSTM, and GRU?

18.Explain the Transformer architecture.

19.What's attention mechanism?

20.What's the vanishing gradient problem?

21.What's the difference between GPT and BERT?

22.Explain how GPT works at a high level.

23.What's fine-tuning vs prompt engineering?

24.What's RAG (Retrieval Augmented Generation)?

25.How would you build a chatbot using LLMs?

26.What's the difference between zero-shot, one-shot, and few-shot learning?

27.What's temperature in LLM sampling?

28.How do you prevent hallucinations in LLMs?

29.What's vector database and when do you need it?

30.What's embedding and how is it used?

31.Explain tokenization.

32.What's RLHF (Reinforcement Learning from Human Feedback)?

33.How would you evaluate an LLM's performance?

34.What's prompt injection?

35.How would you reduce costs when using LLM APIs?

36.Walk me through your ML project from data to deployment.

37.How do you split your data and why?

38.What's your feature engineering process?

39.How do you handle missing data?

40.How do you handle categorical variables?

41.What's data leakage and how do you prevent it?

42.How do you debug a model that's not learning?

43.How do you know when to stop training?

44.What metrics do you use for classification vs regression?

45.How do you deploy an ML model to production?

46.What's model drift and how do you handle it?

47.How do you monitor models in production?

48.What's A/B testing for ML models?

49.How do you handle different versions of models?

50.What's the difference between batch and online inference?

51.Design a recommendation system for Netflix.

52.Design a fraud detection system for a bank.

53.Design a search ranking system.

54.Design a real-time content moderation system.

55.How would you build Instagram's image search?

56.Design a spam detection system for email.

57.How would you build a personalized news feed?

58.Design a chatbot for customer service.

59.How would you scale an ML model to handle millions of requests?

60.Design a system to detect fake news.

61.Implement linear regression from scratch.

62.Write a function to calculate accuracy, precision, recall.

63.Implement k-means clustering.

64.Code softmax function.

65.Implement train-test split.

66.Write code for data normalization.

67.Implement cross-entropy loss.

68.Code a simple neural network in NumPy.

69.Write data preprocessing pipeline.

70.Implement gradient descent.

71.Code for handling imbalanced data.

72.Write evaluation metrics code.

73.Implement confusion matrix.

74.Code for model inference API.

75.Write unit tests for ML code.

76.How do you build a model in PyTorch vs TensorFlow?

77.What's torch.nn.Module?

78.How do you create custom PyTorch Dataset?

79.What's the difference between model.train() and model.eval()?

80.How do you save and load models?

81.Explain PyTorch autograd.

82.What's the difference between tensor.detach() and tensor.data?

83.How do you use Hugging Face Transformers?

84.How do you fine-tune a BERT model?

85.What's tokenizer in Hugging Face?

86.How do you containerize an ML model?

87.What's Docker and why use it for ML?

88.How do you use Kubernetes for ML?

89.What's CI/CD for ML models?

90.How do you track experiments?

91.What tools do you use for model versioning?

92.How do you handle feature engineering in production?

93.What's the difference between training and serving infrastructure?

94.How do you optimize model latency?

95.How do you reduce model size for deployment?

96.What's quantization?

97.How do you handle GPU memory issues?

98.What's batch size and how do you choose it?

99.How do you debug OOM (Out of Memory) errors?

100.How do you profile model performance?

101.How do you work with large datasets that don't fit in memory?

102.What's the difference between SQL and NoSQL for ML?

103.How do you store training data?

104.What's data versioning?

105.How do you create a data pipeline?

106.What's Apache Spark and when do you use it?

107.How do you handle streaming data?

108.What's ETL pipeline?

109.How do you validate data quality?

110.What's feature store?

111.Explain object detection vs image classification.

112.What's YOLO?

113.What's the difference between semantic and instance segmentation?

114.How do you handle different image sizes?

115.What image augmentation techniques do you use?

116.What's transfer learning in computer vision?

117.How would you build a face recognition system?

118.What's U-Net architecture?

119.How do you evaluate object detection models?

120.What's IoU (Intersection over Union)?

121.You have a model with 95% accuracy but it's failing in production. Why?

122.Your model works on training data but fails on test data. What do you do?

123.You need to reduce model inference time by 50%. How?

124.Your dataset has 1,000 positive and 100,000 negative examples. How do you handle this?

125.Your model's performance is degrading over time. What's happening?

126.You need to explain your model to non-technical stakeholders. How?

127.Your training is taking too long. What do you optimize?

128.You have limited labeled data. What techniques do you use?

129.Your model needs to run on mobile devices. What do you do?

130.You're asked to improve an existing model by 5%. What's your approach?

131.Tell me about a challenging ML project you worked on.

132.Describe a time when your model failed in production.

133.How do you stay updated with AI research?

134.Walk me through your ML project from start to finish.

135.What's the biggest mistake you made in an ML project?

136.How do you prioritize features when building models?

137.Describe a time you had to explain ML to non-technical people.

138.What's your approach when you get a new ML problem?

139.How do you handle disagreements about model approach?

140.Why do you want to work in AI?

141.What's matrix multiplication and why is it important?

142.Explain eigenvalues and eigenvectors.

143.What's probability distribution?

144.Explain Bayes theorem.

145.What's derivative and why is it important in ML?

146.What's the chain rule?

147.Explain logarithm and its use in ML.

148.What's exponential function?

149.What's normalization in statistics?

150.What's standard deviation and variance?