Rex.zone

STEM Research Engineer (Canada)

Rex.zone
6 - 10 years
United States
Full-time
Remote
1 month ago

About the role

STEM Research Jobs in Canada
Rex.zone is hiring a STEM Research Engineer to support applied AI/ML research and engineering workflows, including LLM training pipelines, dataset development, RLHF, prompt evaluation, and QA evaluation. This is a remote, full-time role focused on improving training data quality and measurable model performance across NLP and computer vision.
Key Responsibilities
  • Design and execute applied STEM research experiments aligned to model performance improvement and deployment constraints
  • Build and maintain LLM training pipelines combining data labeling, RLHF signals, and QA evaluation outputs
  • Develop prompt evaluation protocols and scoring rubrics for large language model evaluation
  • Implement training data quality checks: annotation guidelines compliance, inter-annotator agreement, and sampling audits
  • Run error analysis for NLP and computer vision tasks (e.g., named entity recognition and annotation edge cases)
  • Support content safety labeling workflows with taxonomies, adjudication rules, and escalation criteria
  • Produce reproducible documentation, experiment logs, and code to ensure traceability
  • Communicate findings and recommendations to technical and non-technical stakeholders

Required Qualifications
  • Mid-Senior experience in STEM research, applied ML research, or engineering research roles
  • Strong Python for data processing, experiment automation, and evaluation tooling
  • Hands-on experience with NLP, computer vision, or multimodal evaluation
  • Knowledge of data labeling, QA evaluation, and annotation guidelines compliance
  • Familiarity with RLHF and prompt evaluation methods
  • Strong experiment design and statistical reasoning
  • Experience collaborating in remote, cross-functional teams

Compensation
$30–$50 per hour (base salary).
Remote, full-time.

Skills

Technology, Information and Internet
See more jobs in United States