Skip to main content
Loading navigation
Back to challenges

Data Center Power & Community Risk Assessor with Claude Agents SDK and Gentrace

Rising political and community backlash against data center power consumption threatens regional grid stability and development permits. Build an autonomous multi-agent evaluation pipeline using Claude Agents SDK and Gentrace to assess site viability, power load constraints, and public opposition risk. Achieve 90% accuracy in predicting municipal permit approval outcomes.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Build an agentic site risk assessment pipeline using Claude Agents SDK instrumented with Gentrace to evaluate power grid load and community sentiment.

How work is evaluated

Evaluation

Evaluates candidate agent accuracy in calculating site risk score and permit recommendation.

Datasets

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

How submissions are scored

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

recommendation_match

Checks if permit recommendation matches ground truth label.

Binary check

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

Dimension 2

risk_score_mae

Mean absolute error of calculated risk score. • target: 5 • range: 0-100

Binary check

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

Learning goals

  • Implement multi-agent analysis using Claude Agents SDK

  • Trace and evaluate AI agent runs using Gentrace pipelines

  • Parse regional power utility capacity and municipal zoning regulations

  • Predict local legislative approval probabilities based on public backlash signals

Resources

Dataset notes

Synthetic dataset of data center site applications including utility grid headroom, power request MW, and public hearing sentiment transcripts.

Challenge setup

Evaluates candidate agent accuracy in calculating site risk score and permit recommendation.

Challenge input

JSON object with site_id, grid_capacity_mw, requested_mw, transcript_text

Claude Agents SDK

Primary framework required for agentic decision making.

Gentrace

Required tracking framework for AI workflow monitoring.

Evaluated output

JSON object with risk_score (0-100), permit_recommendation ('APPROVED'|'REJECTED')

Evaluation checks
  • Checks if permit recommendation matches ground truth label.
  • Mean absolute error of calculated risk score. • target: 5 • range: 0-100
Configured evaluation evidence
  • Benchmark: DataCenterPermitBenchmark
  • Risk Score Mae target: 5
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • RAI · Optional
  • Ellipsis · Optional
Observation
  • Gentrace · Required
Reward / Eval
  • Gentrace · Required

Evaluation contract

  • recommendation_match · Weight 1
  • risk_score_mae · Weight 1

Metrics

  • risk_score_mae · Target 5

Benchmarks

  • DataCenterPermitBenchmark
  • energy

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON object with site_id, grid_capacity_mw, requested_mw, transcript_text

    Output: JSON object with risk_score (0-100), permit_recommendation ('APPROVED'|'REJECTED')

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · Agent Orchestration
  • Gentrace · Evaluation and Tracing

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Data Center Power & Community Risk Assessor with Claude Agents SDK and Gentrace