Agentic Software Test Engineer

trane technologies•La Crosse, WI

2d•$97,407 - $156,200•Onsite

About The Position

At Trane Technologies, the Agentic Software Test Engineer - TRACE is part of the Customer Driven Solutions (C.D.S.) Team and leads the agentic testing organization for TRACE, a building energy simulation platform used by engineers globally for HVAC load calculations, equipment sizing, and ASHRAE-compliant reporting. This role involves directing a team of AI testing agents within a .NET / Blazor codebase, collaborating closely with software developers as a dedicated QA partner. The engineer is responsible for coaching agents, improving quality standards, and ensuring the quality of shipped products. This is a hands-on engineering position requiring direction of agent teams and partnership with human developers.

Requirements

5+ years in software quality assurance, with a real track record of shipping tested software.
1+ year hands-on with LLM-based agents, AI coding assistants, or multi-agent systems in a production or near-production workflow.
Strong test-design instincts across unit, integration, system, functional, and exploratory layers; comfortable owning test plans, defect lifecycle, regression strategy, and release gating.
Hands-on experience with AI coding agents (Claude Code, Cursor, Copilot Workspace, or equivalent) and the Anthropic / OpenAI APIs.
Working knowledge of prompt engineering, structured outputs (JSON Schema), tool use, and multi-agent orchestration (parallel/pipeline/judge-panel/adversarial-verify).
Comfort designing evals and treating eval results as a first-class quality metric.
Bias toward human-in-the-loop verification — never trusting agent output blindly.
Ability to read and modify production code in C#, Python, or JavaScript — not just test code.
Experience with GitHub Actions testing pipelines.
Working knowledge of Jira (or equivalent) and testing tools like Playwright, BUnit, xUnit / NUnit.
Strong written and oral communication; able to explain agent behavior, eval results, and quality posture in plain language.

Nice To Haves

Bachelor's Degree in Engineering, Computer Science, or related discipline (or equivalent experience — we care more about what you've shipped than where you studied).
Familiarity with C#, .NET 10.0, Blazor, ASP.NET Core, Razor, plus the usual web stack (JavaScript / TypeScript / CSS / HTML).
Working knowledge of PostgreSQL and EF Core, including database-aware test design (cascade-delete chains, FK integrity, schema migration testing).
Background in building energy simulation, EnergyPlus, ASHRAE 140 software testing, ASHRAE 90.1 / 62.1 / 169 compliance reporting, or HVAC equipment sizing workflows.
Agile / Scrum experience.

Responsibilities

Direct the agentic testing team — define agent personas, dispatch patterns, and human-in-the-loop checkpoints for QA across TRACE.
Author the prompts, context contracts, and knowledge bases that drive the agents through test plan generation, regression authoring, defect triage, exploratory testing, and accessibility sweeps.
Build evaluation harnesses that measure agent output quality — false-positive rates on bug reports, regression coverage drift, flaky-test detection, eval-vs-production divergence — and tighten the loop sprint over sprint.
Curate what the agents produce — test plans, automated scripts, exploratory findings, regression results — and ship the best of it into the engineering pipeline with your stamp on it.
Hunt bugs through manual, automated, and agent-driven methods, and verify every agent-surfaced finding before it reaches the development team.
Own TRACE's quality posture as the SME on HVAC simulation correctness, ASHRAE compliance reporting, EnergyPlus integration behavior, and cross-repo regression risk.
Partner with developers and Product Owners on test coverage decisions, release readiness, and quality strategy.
Run regression, functional, and data tests across the .NET 10.0 backend, Blazor frontend, PostgreSQL schema, and EnergyPlus calculation engine.
Design adversarial verification workflows — parallel agents that try to refute a test result before it's trusted.
Curate the agent learning loop — capture recurring failure patterns, encode them into spec rules, and feed them back into agent knowledge bases so the same defect class is caught earlier next sprint.
Translate customer-impact reports into agent-runnable test scenarios and explain quality posture to non-technical stakeholders without hand-waving.