AI0-001 AI Implementation and Operations — All Questions With Answers

Refer to the exhibit.

```
Model: ResNet-50
Batch size: 32
Epochs: 10
Learning rate: 0.001
Optimizer: SGD
Data: ImageNet subset
Training accuracy: 0.99
Validation accuracy: 0.75
```

Refer to the exhibit.

```
Error Log:
[2025-03-15 10:23:45] ERROR: Model server 'prod-ml-01' failed health check.
[2025-03-15 10:23:46] WARNING: Inference latency exceeded threshold: 500ms (threshold 200ms).
[2025-03-15 10:23:47] INFO: Rolling restart initiated for 'prod-ml-01'.
```

Refer to the exhibit.

Model: logistic_regression_v1
Features: ['age', 'income', 'loan_amount', 'credit_score']
Training accuracy: 0.87
Test accuracy: 0.85

Deployment metrics (last 24 hours):
  - Accuracy: 0.72
  - Precision: 0.68
  - Recall: 0.81
  - F1: 0.74

Feature distribution shift detected for 'income' (p < 0.05).

Refer to the exhibit.
```
kubectl describe pod model-serving-5f4c6d7b8c-x9yzw
...
Status:       Failed
Reason:       OOMKilled
Message:      The pod was terminated because it exceeded its memory limit.
Containers:
  model:
    Limits:
      memory: 2Gi
    Requests:
      memory: 1Gi
...
```

Refer to the exhibit.
```json
{
  "model_registry": [
    {
      "name": "fraud-detection-v1",
      "version": "1",
      "stage": "Production",
      "artifact_uri": "s3://models/fraud-v1/"
    },
    {
      "name": "fraud-detection-v2",
      "version": "2",
      "stage": "Staging",
      "artifact_uri": "s3://models/fraud-v2/"
    },
    {
      "name": "fraud-detection-v3",
      "version": "3",
      "stage": "Production",
      "artifact_uri": "s3://models/fraud-v3/"
    }
  ]
}
```

Refer to the exhibit.
```
[2024-03-15 10:32:17] ERROR: Exception when invoking model.
  Input tensor shape: [1, 224, 224, 3]
  Model expected shape: [1, 299, 299, 3]
  TensorFlow serving error: Input size mismatch
```

Refer to the exhibit.

```
$ curl -w "@%{http_code}" http://model-serving.example.com/v1/predict -d '{"features": [1.2, 3.4, 5.6]}'
Response: {"prediction": 0.95, "latency_ms": 250}
$ curl -w "@%{http_code}" http://model-serving.example.com/v1/predict -d '{"features": [7.8, 9.0, 1.2]}'
Response: {"prediction": 0.12, "latency_ms": 245}
$ curl -w "@%{http_code}" http://model-serving.example.com/v1/predict -d '{"features": [3.4, 5.6, 7.8]}'
Response: {"error": "Inference timeout", "latency_ms": 500}
```

Refer to the exhibit.

```
{
  "pipeline_version": "2.1",
  "components": {
    "data_ingestion": {
      "source": "s3://data-bucket/transactions/",
      "schedule": "cron(0 2 * * ? *)"
    },
    "feature_engineering": {
      "script": "features.py",
      "parameters": {
        "window_size": 7,
        "aggregation": "mean"
      }
    },
    "model_training": {
      "algorithm": "xgboost",
      "hyperparameters": {
        "n_estimators": 100,
        "learning_rate": 0.1
      },
      "training_data_version": "v1"
    },
    "model_evaluation": {
      "metrics": ["accuracy", "precision", "recall"],
      "threshold": {"accuracy": 0.85, "precision": 0.80}
    },
    "model_deployment": {
      "target": "production",
      "rollback_condition": "if_accuracy_drops_below_0.85"
    }
  }
}
```

Refer to the exhibit.

```
2024-09-17 10:15:23 ERROR Model inference failed: ValueError: Input contains NaN, infinity or a value too large for dtype('float64').
2024-09-17 10:15:23 WARNING Preprocessing step 'normalize' received missing values for feature 'age'.
2024-09-17 10:15:23 INFO Current input row: {'age': nan, 'income': 50000, 'score': 0.75}
2024-09-17 10:15:23 ERROR Batch processing halted after 1000 successful rows.
```

Refer to the exhibit.

$ kubectl get pods
NAME                     READY   STATUS      RESTARTS   AGE
ml-service-7b9c8f-2k4d   0/1     OOMKilled   3          5m
ml-service-7b9c8f-j5p1   1/1     Running     0          10m

$ kubectl logs ml-service-7b9c8f-2k4d
2025/03/15 14:23:45 [FATAL] Out of memory: Killed process 1234 (python)

Refer to the exhibit.

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "sagemaker:CreateModel",
        "sagemaker:CreateEndpointConfig",
        "sagemaker:CreateEndpoint"
      ],
      "Resource": "*"
    }
  ]
}

Refer to the exhibit.

Error: RuntimeError: CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 8.00 GiB total capacity; 6.50 GiB already allocated; 1.50 GiB free; 0 bytes cached) at /workspace/training.py:345

Refer to the exhibit.

```
> show model-monitor
Model: customer_churn_v2
Status: DEPLOYED
Inference: REALTIME
Latency (p99): 250ms
Error Rate: 0.2%
Last Drift Check: 2025-03-15 14:00 UTC
Drift Detected: YES
```

Refer to the exhibit.

```json
{
  "model_policy": {
    "model": "fraud-detection-v3",
    "allowed_roles": ["data_scientist", "ml_engineer"],
    "denied_roles": ["intern"],
    "endpoint": "/api/v1/predict"
  }
}
```