Research Engineer: Automated QC for RL Training Data

Clera; Singapore; On-site; SGD 190,000 - 317,000; Full time

Job summary

Clera is seeking a Research Engineer focused on QC automation for RL training data. The role involves building scalable validation pipelines and auditing systems, partnering with external data vendors, and deriving metrics to assess agent outputs across time zones in a fast-paced environment.

You will implement human-centric QC approaches, avoid overreliance on LLMs, and shape testing criteria for data quality as the platform scales.

Qualifications

• 2–4 years of experience in research engineering or a similar role focused on QC automation.
• Proficiency in Python, Docker, and Linux environments.
• Experience building scalable data validation pipelines and automated QA/QC systems end-to-end.
• Experience working on benchmarks and evaluations for RL training data, defining tasks, rubrics, trajectories.

Responsibilities

• Automate QC for training data created by companies using the platform's infrastructure.
• Build QC systems grounded in human judgment and true understanding, not heavy reliance on LLMs.
• Define and enforce quality standards for RL training data.
• Design experiments and metrics to grade agent outputs.
• Partner with data vendors to debug quality issues, diagnose agent failure modes, and improve data generation processes.
• Translate QC learnings into auditing systems, including sampling strategies and rule-based or model-assisted validation pipelines.
• Continuously integrate QC learnings into infrastructure tools and the vendor portal to reduce anomalies, inconsistencies, and edge cases.

Skills

QC automation

Data validation

Experiment design

Metrics development

Vendor collaboration

Statistics

Communication

Autonomy

Tools

Python

Docker

Linux

Job description

Clera is seeking a Research Engineer focused on QC automation for RL training data. The role involves building scalable validation pipelines and auditing systems, partnering with external data vendors, and deriving metrics to assess agent outputs across time zones in a fast-paced environment.

You will implement human-centric QC approaches, avoid overreliance on LLMs, and shape testing criteria for data quality as the platform scales.