Skip to main content
Back to challenges
Challenge

AI Data Center Microgrid Orchestration using CrewAI and OpenAI o3

This challenge tasks you with managing the energy infrastructure for a 350MW AI data center microgrid using the CrewAI framework. Inspired by Veolia's recent project in Ohio, you will create a crew of autonomous agents designed to balance local renewable generation (solar), backup thermal batteries (Antora Energy style), and grid imports. The agents—including a 'Grid Forecaster', an 'Industrial Load Balancer', and a 'Carbon Optimizer'—must work together to ensure zero downtime for critical AI workloads while minimizing the carbon footprint and operational costs. You will leverage the OpenAI o3 model's advanced reasoning capabilities to handle complex multi-variable optimization. The 'Industrial Load Balancer' agent must dynamically adjust non-critical cooling loads and battery discharge rates in response to solar intermittency and real-time grid pricing. The system must demonstrate resilience by handling simulated 'grid-loss' scenarios where the microgrid must transition to 'island mode' autonomously.

EngineeringHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Design a CrewAI-based autonomous management system using OpenAI o3 to operate a 350MW data center microgrid with solar and thermal storage.

Delivery guide

How work is evaluated

Evaluation

Evaluation is based on the microgrid's ability to maintain a 100% uptime for 'Tier 1' AI loads during a grid outage while maintaining the lowest possible carbon score.

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

Tier 1 Uptime

Critical load must never drop below the required 250MW threshold.

Binary check

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

Dimension 2

Carbon Intensity

Average gCO2 per kWh of energy consumed by the data center. • target: 50 • range: 0-500

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

  • Orchestrate hierarchical agent structures in CrewAI for real-time industrial control

  • Implement OpenAI o3 reasoning loops to solve non-linear energy dispatch problems

  • Model 350MW industrial load profiles including AI server power draw and cooling

  • Develop islanding logic for microgrids using thermal battery discharge strategies

Resources and assets

Reference links and supporting material

Dataset notes

Time-stamped logs of 350MW data center power consumption, solar generation data from Ohio, and synthetic grid frequency telemetry.

Challenge setup

Evaluation is based on the microgrid's ability to maintain a 100% uptime for 'Tier 1' AI loads during a grid outage while maintaining the lowest possible carbon score.

Challenge input

CSV telemetry showing 'grid_status: offline'

CrewAI

Allows specialized agents to collaborate on industrial control tasks.

OpenAI o3

Provides high-level logical reasoning for energy dispatch optimization.

Evaluated output

Log of agent actions and final power balance

Evaluation checks
  • Critical load must never drop below the required 250MW threshold.
  • Average gCO2 per kWh of energy consumed by the data center. • target: 50 • range: 0-500
Configured evaluation evidence
  • Benchmark: MicrogridResilienceBench
  • Carbon Intensity target: 50
  • Protected reference cases support private evaluation
Configured runtime
  • Docker 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
  • OpenAI o3 · Optional
  • CrewAI · Optional
Policy Serving
  • o3 · Required

Evaluation contract

  • Tier 1 Uptime · Weight 1
  • Carbon Intensity · Weight 1

Metrics

  • Carbon Intensity · Target 50

Benchmarks

  • MicrogridResilienceBench
  • energy

Reference cases

  • Protected reference cases are configured.

Workflow tasks

  1. Task 1

    Input: CSV telemetry showing 'grid_status: offline'

    Output: Log of agent actions and final power balance

    Execution harness: Docker

Evaluation tools

  • CrewAI · Agent Orchestration
  • OpenAI o3 · Reasoning Model

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about AI Data Center Microgrid Orchestration using CrewAI and OpenAI o3