Data Scientist - Mathematical Statistics (Python, statsmodels/scipy)

About OpenTrain

OpenTrain is a centralized job board for data-labeling and AI-training work. We aggregate projects from many AI companies and labeling platforms so you can find roles like this one in a single place and apply quickly.

Creating an OpenTrain account is free and makes it easy to discover similar short-term, flexible projects that shape how AI systems behave.

About AI training and this kind of work

AI training (also called data labeling or human feedback work) is the human contribution that helps machine learning models learn and improve. Typical tasks include annotating data, evaluating model outputs, and preparing or validating datasets that models learn from.

This role focuses on statistical analysis to support model evaluation, experiment design, and data-driven decisions. It requires careful interpretation of results, transparent documentation of methods, and reproducible workflows.

The role

You will apply mathematical statistics using Python to analyze and validate datasets used in AI-training projects. This is a part-time contractor position (less than 20 hours per week), open worldwide and paid at $25 per hour.

Expect work with programming-related datasets and evaluation tasks (computer code/programming and rating-style labels). You will use numpy, scipy, statsmodels, and pandas to run analyses, document methods, and communicate findings clearly to stakeholders.

• Employment type: Contractor, Part-time
• Hours: Less than 20 hours/week
• Pay: USD $25 per hour
• Level: Entry level
• Remote: Worldwide

What you'll do

Carry out end-to-end statistical analyses on messy or semi-structured datasets, from cleaning and exploratory work through to final summaries and visualizations.

Design and assess experiments (A/B tests), compute effect sizes and power, fit linear and logistic models, check model assumptions and diagnostics, and construct confidence intervals and p-value–based summaries.

• Clean and wrangle data using pandas and numpy.
• Select and run appropriate statistical tests in scipy/statsmodels (t-tests, chi-square, ANOVA, post-hoc tests).
• Fit and interpret regression models (linear and logistic) and compute/interpret correlation measures (Pearson/Spearman).
• Perform normality checks, use non-parametric methods when needed, and calculate effect sizes and power analyses.
• Produce reproducible notebooks with documented methods, clear narratives, and visual summaries for stakeholders.

Requirements

You must have strong Python skills for data analysis and solid foundations in mathematical statistics. This role relies on your ability to both run appropriate analyses and explain assumptions and limitations clearly.

• Proficiency: numpy, scipy, statsmodels, pandas.
• Mastery of hypothesis testing: t-test, chi-square, ANOVA, and post-hoc procedures when appropriate.
• Ability to calculate and interpret p-values, confidence intervals, and effect sizes.
• Experience with correlation (Pearson/Spearman) and regression modeling (linear and logistic).
• Comfort with probability distributions, normality checks, and non-parametric alternatives.
• Practical experience with data cleaning and exploratory data analysis (EDA).
• Knowledge of experimental design/A-B testing and power analysis.
• Reproducible workflows: notebooks, version control, and clear documentation.
• Strong analytical writing and ability to explain assumptions and limitations.
• SQL for data extraction/joins is a plus.

Who should apply

This project is suited to early-career data scientists, quantitative analysts, or statisticians who are comfortable writing Python analysis code and interpreting statistical results. If you enjoy translating numbers into clear recommendations and producing reproducible analyses, apply.

No specific industry experience is required beyond the listed technical skills, and the role is open to candidates worldwide as long as you can work independently and communicate in clear written English.

How it works

Assignments will come as short tasks or mini-projects. You will submit reproducible notebooks and documentation that show your code, assumptions, diagnostics, and final summaries. Platforms and tooling may vary between projects.

OpenTrain aggregates the posting and application process; once selected you will work under a contract and follow the project's specific submission and review workflow. Preserve transparency in your methods so results can be checked and reused.

• Data type: computer code/programming datasets and evaluation labels.
• Labeling/analysis software: varies (OTHER).
• Deliverables typically include notebooks, written summaries, and visualizations.