Data scripts & ETL

Build data processing pipelines with Claude

Data processing is a Claude Code sweet spot

Writing scripts to process CSVs, transform data, move it between systems, or generate reports is tedious, error-prone manual work. Claude Code handles it well — it can read your actual data files, understand the structure, write the processing logic, and test it against real data.

CSV processing workflow

"I have a CSV at /data/products.csv. Read the first 5 rows and tell me what the columns are and what the data looks like."

[Claude reads and reports the structure]

"Now write a script that:
1. Reads the CSV
2. Filters to rows where status = 'active' and price > 0
3. Transforms the 'product_name' field to title case
4. Exports to /data/products_clean.csv

Use Node.js with the csv-parse library. Handle encoding issues gracefully — this CSV came from Excel."

A simple ETL pattern

"Build a simple ETL script that:
EXTRACT: Reads from [source]
TRANSFORM: [describe the transformations]
LOAD: Writes to [destination]

Requirements:
- Log each step with timestamp
- Handle errors gracefully (don't crash on one bad record)
- Show a summary at the end: X records processed, Y errors
- Dry-run mode that shows what would happen without writing anything"

Data validation

"Before we import this client data, validate it:
- Every row must have: name, email, phone
- Email must be valid format
- Phone must be 10 digits (US)
- No duplicate emails

Read /data/client_import.csv, run validation, and report:
- Total rows, Valid rows, Invalid rows with the specific reason for each"