Cloud Infrastructure Cost Triage Agent with Google ADK and Mastra AI
Managing high GPU infrastructure costs during cloud expansion presents strict fiscal and operational challenges for data platforms. Build an intelligent cost triage agent using Google ADK with Gemini and Mastra AI workflow nodes to identify underutilized compute resources, analyze pricing paradoxes, and automate dynamic instance right-sizing.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct a Google ADK agent integrated with Mastra AI tools to monitor sovereign cloud GPU usage metrics and issue automated right-sizing recommendations.
How work is evaluated
Evaluates Google ADK agent recommendations against benchmark cloud telemetry workloads.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
correct_underutilized_identification
Verifies that instances with <10% utilization are correctly targeted.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
cost_saving_efficiency
Ratio of accurately identified savings vs total potential waste. • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Use Google Agent Development Kit (ADK) to build Gemini-powered infrastructure optimization agents.
Integrate Mastra AI workflow constructs for orchestration across heterogeneous cloud telemetry.
Analyze GPU utilization metrics to resolve pricing paradox issues in public cloud DPI setups.
Reference links and supporting material
Telemetry logs from cloud infrastructure containing node GPU/CPU memory utilization, hourly pricing, and active container assignments.
Challenge setup
Evaluates Google ADK agent recommendations against benchmark cloud telemetry workloads.
Challenge input
JSON containing cluster telemetry, hourly costs, and utilization percentage.
Google ADK
Native agent framework for Gemini integration
Mastra AI
Modular agent step orchestrator
Evaluated output
JSON containing action_plan, estimated_cost_reduction_pct, and target_instances.
- Verifies that instances with <10% utilization are correctly targeted.
- Ratio of accurately identified savings vs total potential waste. • target: 0.95 • range: 0-1
- Benchmark: Cloud FinOps Agent Benchmark
- Cost Saving Efficiency target: 0.95
- 1 public reference case
- 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
- Google ADK · Required
- Mastra AI · Optional
- Google · Optional
- Mastra AI · Optional
Evaluation contract
- correct_underutilized_identification · Weight 1
- cost_saving_efficiency · Weight 1
Metrics
- cost_saving_efficiency · Target 0.95
Benchmarks
- Cloud FinOps Agent Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing cluster telemetry, hourly costs, and utilization percentage.
Output: JSON containing action_plan, estimated_cost_reduction_pct, and target_instances.
Execution harness: Python
Evaluation tools
- Google ADK · agent_framework
- Mastra AI · agent_systems
Recipe freeze marker
No recipe freeze time is recorded.