Numeric Data Preprocessing for ML

I’m sitting on large, messy numeric datasets and need them cleaned, engineered, and split so downstream models train smoothly. Your mission is to design a reproducible preprocessing pipeline: audit the raw CSVs, tackle missing or extreme values, scale or normalize where it counts, generate meaningful features, and output train/validation/test sets ready for model development.

Python with pandas, NumPy, and scikit-learn is my default tech stack, but feel free to weave in any modern data-ops tricks—as long as the final workflow runs inside a Docker container that can drop into our CI pipeline.

Deliverables
• Clean, documented datasets (CSV or Parquet)
• Version-controlled preprocessing scripts/notebooks
• Dockerfile plus a concise README for one-command reproduction
• Short report summarizing decisions, assumptions, and next-step suggestions

Send a detailed project proposal outlining your approach, timeline, and anything from previous jobs that proves you thrive with large numeric data. I’ll shortlist based on the clarity and completeness of that plan. Let’s make this data model-ready.