Human Data Quality Analyst, AI Business

Role overview Join the human-data side of frontier AI as an early-career analyst focused on data quality for AI training and evaluation. The work goes beyond traditional QA, combining statistical analysis with close review of human-generated annotations to identify failure modes and drive improvements across programmes. It is an analytical role designed for someone looking to grow quickly while contributing to a fast-evolving AI infrastructure platform.

Responsibilities

- Run day-to-day quality measurement across live human-data programmes using statistical analysis and hands-on review. - Investigate quality shifts, quantify impact, and distinguish individual errors from systemic issues in guidance, task design, or tooling. - Help build the data-quality pipeline alongside quality engineers, including rule-based, LLM-assisted, and automated validations. - Translate complex analysis into clear reporting and visualisations for technical and non-technical stakeholders. - Support new programmes from design through launch, contributing to rubrics, guidelines, gold sets, and calibration exercises. - Use findings from live programmes to improve annotation design, guidance, training, and overall quality processes.

Requirements

- One to two years of practical experience analysing real data in an analytical, quality, research, or data-focused role. - Working proficiency in SQL and Python for queries, analysis, and interpreting results. - Applied understanding of statistical concepts such as sampling, distributions, and variability. - Ability to turn evidence into clear accounts with concrete examples and practical recommendations. - Careful, consistent assessment of detailed work combined with pattern recognition across datasets. - Comfort testing approaches, documenting reasoning, and asking for input when the right method is unclear.

Nice to have

- Experience with data produced or judged by people, such as annotation, labelling, research coding, or survey research, including agreement or gold-set metrics. - Background in psychology, behavioural science, linguistics, or a related field. - Familiarity with how human data supports supervised fine-tuning, RLHF, preference data, and evaluation benchmarks. - Experience creating dashboards or clear data visualisations.

Benefits and work setup

- Hybrid working model with access to a unique human-data platform used by frontier AI researchers. - Compensation packages include base salary, equity, and benefits; many roles also include a bonus or commission element.