Skip to main content
Back to challenges
Challenge

Human-AI Collaborative Data Engineering Pipeline

Data engineering teams require active human oversight when AI agents transform public datasets. Create an automated human-in-the-loop pipeline that cleans public datasets, flags data quality anomalies, and routes uncertain transformations to human engineers.

Workflow AutomationHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Develop an interactive human-in-the-loop data pipeline for automated data cleaning and anomaly triage.

Delivery guide

How work is evaluated

Evaluation

Evaluates pipeline ability to flag ambiguous records for human review vs auto-processing clear records.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

correct_triage

Verifies that invalid age triggers human review

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

triage_precision

Precision of routing corrupt data to human review • target: 0.95 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Build dynamic data quality profiling scripts

  • Implement human reviewer decision handoff interfaces

  • Automate target ETL loads post-human verification

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic public service request records containing intentional noise and edge cases.

Challenge setup

Evaluates pipeline ability to flag ambiguous records for human review vs auto-processing clear records.

Challenge input

JSON dataset record with missing or corrupted fields

Great Expectations

Automated data validation expectations

Evaluated output

JSON containing action (AUTO_CLEAN/HUMAN_REVIEW) and confidence_score

Evaluation checks
  • Verifies that invalid age triggers human review
  • Precision of routing corrupt data to human review • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: DataQuality-Bench
  • Triage Precision target: 0.95
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Evidently AI · Required
  • Shakudo · Optional
  • Haystack · Optional

Evaluation contract

  • correct_triage · Weight 1
  • triage_precision · Weight 1

Metrics

  • triage_precision · Target 0.95

Benchmarks

  • DataQuality-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON dataset record with missing or corrupted fields

    Output: JSON containing action (AUTO_CLEAN/HUMAN_REVIEW) and confidence_score

    Execution harness: Python

Evaluation tools

  • Great Expectations · Data Quality Tool

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Human-AI Collaborative Data Engineering Pipeline