Job description

About the role Our client is looking for experienced data scientists to evaluate how frontier AI models handle real data science work: an analysis from messy data to a defensible conclusion, an experiment designed and read out correctly, a model built and validated without leakage, SQL and Python that actually does what the ticket asked. You bring the judgment you have built sending back a notebook that passed its cells but not its logic, and knowing the difference between a result and a real result. We bring the model output that judgment is needed to grade. In this role, you will design challenging, realistic tasks drawn from your own practice, such as an exploratory analysis and write up from a messy dataset, an A/B test design and read out, a feature set and model with validation, a SQL pipeline with data quality checks, a metric definition and dashboard spec, or a forecast with error analysis, run them through frontier AI agents, and evaluate what comes back against a professional standard. You will work with realistic professional files, the kind a practitioner in your field actually handles, which you assemble yourself. Some tasks are compact, built around a handful of files; others are larger scenarios that take several days to build. In every case the goal is the same: a task a competent professional in your field would complete correctly and a frontier model currently gets wrong. This is not a traditional data science role. You will be helping build better AI by putting your knowledge to work in a structured, flexible, fully remote environment. The work is long form and self directed, and clear written reasoning matters as much as technical depth. Responsibilities • Design challenging, realistic data science tasks drawn from your own day to day work: the scenario, a prompt phrased the way you would brief a trusted colleague, and the supporting files a practitioner would need (datasets you construct or anonymize, schemas, notebooks, metric definitions, stakeholder briefs), which you author yourself. • Run those tasks through frontier AI models and evaluate the deliverable they produce (the analysis, code, model, query or write up) against the standard you would hold a colleague to. • Compare two model outputs on identical prompts and files, decide which performed better, and document where each fell short. • Write detailed grading rubrics and, where appropriate, tests that specify what a correct deliverable must contain, such as the right handling of the data quality problem, the right statistical test, the right validation approach and the right interpretation, and explain in writing why a response passes or fails each one. • Flag concrete failures with evidence: leakage, confounding, wrong joins, misapplied tests, code that runs but does not do what was asked, fabricated or ignored data, and off brief interpretation of the ask. • Contribute across analytics, machine learning, experimentation and data engineering adjacent work, and review and refine tasks built by other experts. Domain qualifications • 2+ years of professional applied data science experience preferred (analytics, machine learning, statistical modeling, experimentation, analysis adjacent data engineering) in industry or research. Undergraduate study does not count toward the experience floor. • In progress Bachelor’s degree or higher in a quantitative field, completed or in progress. Advanced degree is a plus but optional; practical applied work outweighs credentials; cloud and analytics certifications are neither required nor screened for. • Working proficiency in Python and/or SQL. You write, debug and can explain your own analysis code; reading and writing code is part of the work. • Depth in at least one of: product or business analytics and BI; applied machine learning; statistics and experimentation (A/B testing, causal inference); forecasting; analysis adjacent data engineering; applied NLP or computer vision with shipped work. • Has independently owned a multi step analysis end to end, from raw or messy data through cleaning, modeling and interpretation to a decision, rather than picking up a clean dataset mid pipeline. • Can judge whether a result is real and say what it means: leakage, confounding, selection effects, multiple testing, power, and interpreting results in business or scientific terms. General requirements • 2+ years of hands on experience in your field preferred (see Domain qualifications above). Candidates with less experience are considered where the practical work is real. • Able to draw on your own real world experience and day to day workflows to craft scenarios that test whether an AI system can actually do the work. • Hands on practitioner: you currently do (or recently did) the work yourself at an individual contributor level, not solely in a managerial capacity. • Full professional or native level written and spoken English, with strong written communication. You can explain complex professional reasoning clearly and concisely, and articulate why a result is wrong, not only that it is. • Comfort with ambiguity and attention to detail. You can orient in a new set of files and build an accurate, deep working picture of it quickly, especially when the subject sits partly outside your own specialization. You verify what a document claims against the underlying numbers, sources or facts. • Capable of interpreting feedback, judging which parts of it are actually correct, and applying it without hand holding. When stuck, you look for the answer rather than waiting for one. • Ability to ramp quickly on unfamiliar work from written material and instructions alone, including where that material is incomplete (for example, writing grading rubrics for the first time). • General familiarity with AI and LLM tools. You have used models like Claude or ChatGPT in professional work and have the judgment to tell a well reasoned answer from a plausible sounding but incorrect one. • Baseline tech literacy: comfortable with cloud file tools (e.g., Google Workspace), managing browser profiles, downloading and installing desktop apps (e.g., Claude), and everyday file handling (e.g., converting between Excel and Google Sheets, zipping files for sharing). • Available at least 10 hours per week, with no weekly maximum. Consistent availability is valued and full time hours are available. • Based in the United States, Canada, or the UK.

Requirements

Must-haves

Run those tasks through frontier AI models and evaluate the deliverable they produce (the analysis, code, model, query or write up) against the standard you would hold a colleague to.
Compare two model outputs on identical prompts and files, decide which performed better, and document where each fell short.
Flag concrete failures with evidence: leakage, confounding, wrong joins, misapplied tests, code that runs but does not do what was asked, fabricated or ignored data, and off brief interpretation of the ask.
Contribute across analytics, machine learning, experimentation and data engineering adjacent work, and review and refine tasks built by other experts.
Domain qualifications
Working proficiency in Python and/or SQL. You write, debug and can explain your own analysis code; reading and writing code is part of the work.
Has independently owned a multi step analysis end to end, from raw or messy data through cleaning, modeling and interpretation to a decision, rather than picking up a clean dataset mid pipeline.
Can judge whether a result is real and say what it means: leakage, confounding, selection effects, multiple testing, power, and interpreting results in business or scientific terms.
General requirements
Able to draw on your own real world experience and day to day workflows to craft scenarios that test whether an AI system can actually do the work.
Hands on practitioner: you currently do (or recently did) the work yourself at an individual contributor level, not solely in a managerial capacity.
Capable of interpreting feedback, judging which parts of it are actually correct, and applying it without hand holding. When stuck, you look for the answer rather than waiting for one.
Ability to ramp quickly on unfamiliar work from written material and instructions alone, including where that material is incomplete (for example, writing grading rubrics for the first time).
General familiarity with AI and LLM tools. You have used models like Claude or ChatGPT in professional work and have the judgment to tell a well reasoned answer from a plausible sounding but incorrect one.
Available at least 10 hours per week, with no weekly maximum. Consistent availability is valued and full time hours are available.
Based in the United States, Canada, or the UK.

Nice-to-haves

2+ years of hands on experience in your field preferred (see Domain qualifications above). Candidates with less experience are considered where the practical work is real.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aisqlpythonmachine learningllm

Platform: wp-job-manager
Effort: medium

Benefits

Benefits detected in the listing.

Vision