Skip to main content
Back to challenges
Challenge

Mastra AI Auto-scaling Inference Fleet with Thermal Constraints

Thermal throttling on dense GPU clusters severely degrades token generation performance during long-running tasks. Build a Mastra AI TypeScript workflow that monitors GPU junction temperatures and power usage, executing dynamic fleet autoscaling and workload migration before thermal degradation occurs.

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.

Develop a Mastra AI agent workflow that monitors accelerator thermal limits and dynamically migrates inference workloads across cloud nodes.

Delivery guide

How work is evaluated

Evaluation

Evaluates thermal-aware autoscaling and migration logic during temperature spike simulation.

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

thermal_safety_guarantee

Ensures workloads were migrated off nodes exceeding 82°C.

Binary check

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

Dimension 2

migration_latency_ms

Time taken to execute workload reassignment • target: 80 • range: 10-500

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

  • Construct a Mastra AI workflow with dynamic step execution and telemetry inputs

  • Monitor accelerator junction temperature (°C) and fan speed trends

  • Implement proactive workload migration rules to idle nodes upon reaching 82°C threshold

  • Autoscale cluster replica count under high queue depth while preserving thermal headroom

Resources and assets

Reference links and supporting material

Dataset notes

Thermal telemetry streams recording GPU core and memory junction temperatures under continuous load.

Challenge setup

Evaluates thermal-aware autoscaling and migration logic during temperature spike simulation.

Challenge input

JSON containing node_telemetry, cluster_capacity, active_workloads

Mastra AI

Open-source TypeScript framework for workflows and agents

Evaluated output

JSON containing migration_triggered, target_node, migrated_workload_count, cluster_thermal_avg_c

Evaluation checks
  • Ensures workloads were migrated off nodes exceeding 82°C.
  • Time taken to execute workload reassignment • target: 80 • range: 10-500
Configured evaluation evidence
  • Benchmark: Data Center Energy Efficiency Standard
  • Migration Latency Ms target: 80
  • 1 public reference case
Configured runtime
  • JavaScript execution harness
  • javascript 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
  • Mastra AI · Required
  • Anthropic · Optional
  • Anthropic Claude · Optional
Orchestration
  • Mastra AI · Required

Evaluation contract

  • thermal_safety_guarantee · Weight 1
  • migration_latency_ms · Weight 1

Metrics

  • migration_latency_ms · Target 80

Benchmarks

  • Data Center Energy Efficiency Standard
  • energy

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing node_telemetry, cluster_capacity, active_workloads

    Output: JSON containing migration_triggered, target_node, migrated_workload_count, cluster_thermal_avg_c

    Execution harness: JavaScript

Evaluation tools

  • Mastra AI · TypeScript Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Mastra AI Auto-scaling Inference Fleet with Thermal Constraints