AI Prompts for Gemini for Data Science

I have a dataset with these columns and data types: [DESCRIBE OR PASTE SAMPLE]. What exploratory analysis should I run first, and write the Python / pandas code to do it?

This dataset has quality issues: [DESCRIBE ISSUES, E.G., MISSING VALUES, OUTLIERS, INCONSISTENT FORMATS]. Write the pandas code to identify and handle each issue, explaining the trade-offs of each approach.

Write a data profiling script for this dataset: [DESCRIBE STRUCTURE]. I want to see: null counts, value distributions, unique counts, and basic statistics for each column.

I need to merge these two datasets: [DESCRIBE DATASET A AND DATASET B]. The join key is [KEY]. What are the potential join issues (duplicates, nulls, type mismatches) and how should I handle them?

My dataset has [N] features. I need to select the most relevant features for predicting [TARGET]. What feature selection approaches should I use, and write the code for each.

I want to predict [TARGET] from [FEATURES]. My data: [DESCRIBE SIZE AND TYPE]. What model should I start with and why? Write the full scikit-learn pipeline including preprocessing.

Explain the trade-offs between [MODEL A] and [MODEL B] for my use case: [DESCRIBE PROBLEM AND DATA]. Which should I use and what hyperparameters matter most?

Write Python code to tune the hyperparameters of [MODEL] using [GRIDSEARCHCV / RANDOMIZEDSEARCHCV / OPTUNA]. Metric: [METRIC]. Include cross-validation and a summary of the best parameters.

My model has [DESCRIBE ISSUE: HIGH VARIANCE / LOW ACCURACY / SLOW INFERENCE / CLASS IMBALANCE]. Diagnose the likely cause and suggest the highest-impact fixes.

Write a model evaluation function that computes [METRICS: ACCURACY, F1, AUC-ROC, RMSE, ETC.] for [CLASSIFICATION / REGRESSION] and produces a clean summary report with visualizations.

Write Python code to visualize [DESCRIBE WHAT YOU WANT TO SHOW] using [MATPLOTLIB / SEABORN / PLOTLY]. The audience is [TECHNICAL / NON-TECHNICAL]. Make it clear and presentation-ready.

I have this analysis result: [DESCRIBE FINDING]. Write a plain-language explanation for a business stakeholder who is not technical. Focus on the so-what, not the methodology.

Create a data dashboard in [GOOGLE LOOKER STUDIO / COLAB / PLOTLY DASH] for [DESCRIBE WHAT TO TRACK]. Include: key metrics, trend charts, and filter controls.

Write a data story for this dataset: [DESCRIBE OR PASTE SAMPLE FINDINGS]. Structure it as: the question, the data, the key insight, and the recommendation.

Generate a Jupyter notebook structure for a data science project on [TOPIC]. Include: section headers, markdown cells explaining each stage, and placeholder code cells.

Write Python code to query BigQuery and load the results into a pandas DataFrame for analysis. Query: [DESCRIBE WHAT DATA YOU NEED]. Handle authentication and pagination.

I have a large dataset in Google Cloud Storage. Write Python code to load it efficiently for analysis: [DESCRIBE FORMAT, E.G., PARQUET / CSV / JSON]. Handle memory constraints.

Set up a data science environment on Google Colab for [PROJECT TYPE]. What libraries should I install, how should I authenticate to Google Cloud, and how should I structure my notebooks?

Write Python code to train a model and deploy it to Vertex AI. Model: [DESCRIBE]. Input/output schema: [DESCRIBE]. Include: training script, model export, and deployment endpoint creation.

I want to use BigQuery ML to [DESCRIBE ML TASK]. Write the BigQuery SQL to create, train, and evaluate the model. Explain when to use BQML vs. Python-based modeling.