The work
You will help test AI models with realistic HVAC service and system design problems. Your work will turn practical trade knowledge into examples that show how symptoms lead to diagnosis, repair, and safe system decisions.
You will also review model answers and identify advice that is technically wrong, inefficient, or dangerous. This includes spotting recommendations that could damage a compressor or introduce carbon monoxide into a home.
- Create HVAC scenarios that move from symptoms through diagnosis and repair.
- Write reference answers covering system behavior, charge selection, duct sizing, and sequence of operations.
- Evaluate model responses for technical accuracy, efficiency, and safety.
- Identify plausible but unsafe or incorrect HVAC recommendations.
What it pays and takes
This is remote, asynchronous contractor work. The role is open to candidates in the countries listed for the project, including the United States, Canada, the United Kingdom, Australia, and many other countries across several regions.
The role requires practical HVAC knowledge and the ability to explain diagnostic reasoning clearly in written English. Additional experience in residential service, commercial refrigeration, controls, heat pumps, geothermal systems, high-efficiency systems, building automation, or NATE certification is helpful.
- Schedule: At least 20 hours per week, with no maximum weekly commitment stated.
- Payment: Paid weekly via Stripe.
- Certification: EPA Section 608 certification required.
- Licensing: State HVAC licensing is required where applicable.
- Experience: At least two years in HVAC installation, service, or controls.
- Knowledge: Refrigeration cycles, airflow, and load calculation.
- Language: Clear written English for documenting technical reasoning.
How it works
Apply on OpenTrain with your resume, then complete the application on the hiring site.
About AI training work
AI training is the human work behind systems that learn from examples and feedback. In this role, your HVAC judgment helps evaluate whether an AI model gives answers that are accurate, efficient, and safe.