Skip to main content
Back to challenges
Challenge

Smart Building HVAC Telemetry Energy Balancing Agent

Commercial buildings represent a primary application ground for autonomous agent systems. Construct an energy balancing control agent using FastAI and PyTorch to dynamically adjust HVAC air delivery rates based on occupancy telemetry and outdoor weather profiles. Reduce simulated cooling kilowatt-hour consumption by 12%+ without compromising comfort parameters.

Business OperationsHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Build an intelligent HVAC control loop using FastAI neural networks to optimize energy consumption against thermal comfort bounds.

Delivery guide

How work is evaluated

Evaluation

Evaluates energy reduction percentage and thermal comfort threshold adherence.

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

Comfort Boundary Guardrail

Indoor temperature must remain between 70F and 75F

Binary check

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

Dimension 2

Energy Savings Percentage

Percentage reduction in cooling kWh compared to constant 100% baseline • target: 12 • range: 5-25

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

  • Train tabular neural network models in FastAI on building energy telemetry

  • Predict thermal inertia and indoor ambient temperature decay rates

  • Formulate control loss functions balancing kWh consumption against target temperatures

  • Deploy optimized control policy loop in Python runtime

Resources and assets

Reference links and supporting material

Dataset notes

Hourly HVAC fan speeds, occupant counts, indoor temperatures, and kWh draw from commercial office buildings.

Challenge setup

Evaluates energy reduction percentage and thermal comfort threshold adherence.

Challenge input

JSON with occupancy_count, outdoor_temp_f, current_indoor_temp_f

FastAI

Provides fast tabular neural network training primitives

Evaluated output

JSON with optimal_fan_speed_pct, projected_kwh_reduction_pct, comfort_violated

Evaluation checks
  • Indoor temperature must remain between 70F and 75F
  • Percentage reduction in cooling kWh compared to constant 100% baseline • target: 12 • range: 5-25
Configured evaluation evidence
  • Benchmark: HVACControlBench
  • Energy Savings Percentage target: 12
  • 1 public reference case
Configured runtime
  • 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

Action Space
  • Synthflow · Optional
  • Google ADK · Optional
Observation
  • Portkey · Required
Policy Serving
  • Portkey · Required

Evaluation contract

  • Comfort Boundary Guardrail · Weight 1
  • Energy Savings Percentage · Weight 1

Metrics

  • Energy Savings Percentage · Target 12

Benchmarks

  • HVACControlBench
  • energy

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with occupancy_count, outdoor_temp_f, current_indoor_temp_f

    Output: JSON with optimal_fan_speed_pct, projected_kwh_reduction_pct, comfort_violated

    Execution harness: Python

Evaluation tools

  • FastAI · Deep Learning Library

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Smart Building HVAC Telemetry Energy Balancing Agent