Research Engineer, QC Automation

HUD builds infrastructure for companies creating training data for AI agents. As demand grows, we need robust systems that can maintain and scale data quality. As a Research Engineer, QC Automation , you’ll build the systems that make that possible. You’ll develop automated quality-control infrastructure grounded in human judgment and a deep understanding of what makes training data useful-not simply by relying on LLMs to judge other LLMs. This is a highly autonomous role for someone who enjoys ambiguous problems, learns quickly, and can turn unclear quality requirements into measurable, scalable systems. What You’ll Do

• Build quality-control systems grounded in human judgment and a deep understanding of data quality, with limited reliance on LLM-based evaluation.
• Define, formalize, and enforce quality standards for training data.
• Design experiments, benchmarks, and metrics to evaluate agent performance.
• Partner with data vendors to identify agent failure modes , debug quality issues, and improve data-generation processes.
• Build systems for auditing supplier datasets , including sampling strategies, rule-based validation, model-assisted validation, and feedback loops.
• Integrate QC insights into HUD’s infrastructure and data-vendor portal to reduce anomalies, inconsistencies, and edge cases.
• Work across unfamiliar domains and turn loosely defined quality problems into reliable, repeatable processes.

Experience: Technical aptitude and learning potential matter more than years of experience., + Strong proficiency in Python, Docker, and Linux environments.

• Excellent judgment about what constitutes good data and how to measure it.
• Genuine curiosity about unfamiliar domains, with the ability to ask the right questions to develop a deep understanding quickly.
• Experience building scalable data-validation pipelines, automated QA/QC systems, or similar infrastructure without a prescribed roadmap.
• Experience designing or working with benchmarks and evaluations.
• Comfort working in an early-stage startup environment , taking ownership and executing independently.

• A track record of learning quickly and tackling problems that don#39;t come with clear instructions. Strong Signals

• Knowledge of statistics and experimental design.
• Strong written and verbal communication skills.
• Comfort designing metrics, experiments, and QA/QC processes.
• Ability to construct tasks for new evaluations and benchmarks.
• Comfort operating in unstructured problem spaces.
• A tendency to dig into the underlying problem rather than reaching for the first available tool or solution.