Process text, image, audio, and tabular data

Different data types require different preprocessing.

• cleaning
• normalization
• segmentation
• metadata extraction

• format validation
• resizing if required
• OCR / visual understanding
• metadata extraction

• transcription
• language detection
• normalization

• schema validation
• missing-value handling
• type normalization
• transformation

• Amazon Bedrock multimodal models
• Amazon SageMaker Processing
• Amazon Transcribe
• AWS Lambda
• Amazon S3

                ┌── Text Processing ──┐
Raw Data ───────┼── Image Processing ┤
                ├── Audio → Transcribe
                └── Tabular Processing
                         ↓
                  Normalized Data
                         ↓
                          FM

Do not assume every FM accepts every data type in the same format.

---