A
Claude Code logoClaude CodeRun reproducible ML experiments

ML Experiment Lab

Setuproll editorial@setuproll
89.0Overall score

A data and ML setup that explores a dataset, engineers features, and tracks every run with seeds and metrics so results stay trustworthy. For practitioners who want an agent that guards against leakage instead of overstating accuracy.

89.0Score
5Components

Install this build

Export
terminal
pip install jupyterlab mlflow && jupyter lab

Components

Model

  • Claude Opus 4.8

MCP servers

  • jupyter
  • postgres
  • filesystem
  • github

Subagents

  • eda-explorer
  • feature-engineer
  • eval-reviewer

Hooks

  • PostToolUse: run cell and capture metrics
  • Stop: log experiment to tracker

Rules

  • Seed everything for reproducibility
  • Never leak the test set into features

Summary

A data and ML setup that explores a dataset, engineers features, and tracks every run with seeds and metrics so results stay trustworthy. For practitioners who want an agent that guards against leakage instead of overstating accuracy.

89.0 score

0 Reviews

Your rating
Sign in to post

Loading discussion...