senior-data-scientist

What This Skill Does

The Senior Data Scientist skill is a comprehensive toolkit designed for production-level statistical modeling, causal inference, and machine learning pipelines. It empowers OpenClaw to act as an expert collaborator capable of designing rigorous A/B tests, conducting feature engineering, and evaluating model performance using industry-standard metrics. Whether you are dealing with structured tabular data or complex observational studies, this skill bridges the gap between raw data and actionable business intelligence. It provides pre-validated methodologies for sample size calculation, difference-in-differences analysis, and cross-validated predictive analytics using a robust stack of Python libraries like Scikit-learn, NumPy, and Pandas.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/alirezarezvani/senior-data-scientist Ensure you have the necessary environment dependencies installed to support advanced scientific Python packages.

Use Cases

• Experimentation: Design and analyze A/B tests with strict statistical controls, including MDE (Minimum Detectable Effect) calculation, power analysis, and Bonferroni corrections for multiple testing.
• Machine Learning Pipelines: Build automated feature engineering workflows using ColumnTransformers and Pipelines to ensure consistency between training and inference environments.
• Causal Analysis: Perform observational studies and difference-in-differences analysis to uncover insights when controlled experiments are not feasible.
• Model Evaluation: Conduct thorough performance audits using AUC-ROC, AUC-PR, and interpretability tools like SHAP to ensure model fairness and reliability.
• Business Intelligence: Translate complex statistical outputs—such as confidence intervals, lift metrics, and feature importance—into clear, data-driven recommendations for stakeholders.

Example Prompts

1. "I am planning a website conversion experiment with a 10% baseline. Calculate the required sample size per variant for a 5% relative MDE at 80% power."
2. "Review my feature engineering pipeline for this churn prediction model. I need to impute missing values, scale numeric columns, and one-hot encode categorical features."
3. "The A/B test results are in. Please perform a two-proportion z-test on these conversion numbers and provide the lift, confidence interval, and interpretation."

Tips & Limitations

• Data Integrity: Always ensure your data is clean before feeding it into the pipeline; garbage in, garbage out applies heavily here.
• Sample Ratio Mismatch (SRM): Always check for SRM before analyzing experiments to ensure data collection logs are accurate.
• Business Logic: While the models provide statistical significance, always layer in domain expertise to validate if the findings make sense for your specific business cycle.
• Multi-tasking: Remember to manually apply the Bonferroni correction if your objective is to monitor more than one primary outcome metric to avoid the p-hacking fallacy.