Skip to main content
Loading navigation
Back to challenges

Built Environment Predictive HVAC Anomaly Assistant with Google ADK

Commercial facilities require autonomous monitoring to cut energy waste and prevent system outages. Construct a multimodal predictive maintenance assistant using Google's Agent Development Kit (ADK) and Gemini models to process building acoustic recordings and sensor telemetry. Achieve 90%+ diagnostic accuracy on motor and compressor faults.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Build an intelligent building maintenance assistant using Google ADK and Gemini multimodal processing to diagnose HVAC anomalies.

How work is evaluated

Evaluation

Evaluates acoustic and sensor fault classification precision using Google ADK pipelines.

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

Fault Classification Test

Accurately flags high-pitch grinding as bearing failure

Binary check

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

Dimension 2

Diagnostic Accuracy

Accuracy across multimodal fault test set • target: 0.92 • range: 0.85-1

Binary check

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

Learning goals

  • Initialize Google ADK agents with Vertex AI model support

  • Combine audio inspection files with time-series temperature telemetry

  • Identify bearing wear and compressor cavitation anomalies

  • Publish structured maintenance action plans via Vertex AI tools

Resources

Dataset notes

Multimodal dataset of commercial HVAC compressor acoustic clips and telemetry time series.

Challenge setup

Evaluates acoustic and sensor fault classification precision using Google ADK pipelines.

Challenge input

JSON with compressor_vibration_hz, motor_temp_c, acoustic_label

Google ADK

Official framework for building Google Gemini agents

Evaluated output

JSON with predicted_fault, recommended_action

Evaluation checks
  • Accurately flags high-pitch grinding as bearing failure
  • Accuracy across multimodal fault test set • target: 0.92 • range: 0.85-1
Configured evaluation evidence
  • Benchmark: FacilityDiagBench
  • Diagnostic Accuracy target: 0.92
  • 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
  • Google · Required
  • Google ADK · Optional
  • Google Gemini · Optional

Evaluation contract

  • Fault Classification Test · Weight 1
  • Diagnostic Accuracy · Weight 1

Metrics

  • Diagnostic Accuracy · Target 0.92

Benchmarks

  • FacilityDiagBench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with compressor_vibration_hz, motor_temp_c, acoustic_label

    Output: JSON with predicted_fault, recommended_action

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Built Environment Predictive HVAC Anomaly Assistant with Google ADK