Churn Prediction with Sphinx - Coding AI Agent for Data Scientists in VS Code

Описание: In this video, I try out Sphinx, an AI coding agent for solving data science tasks directly inside Visual Studio Code. I start with an empty Jupyter notebook and give it a simple prompt: “Build a classifier for customer churn.”

From there, Sphinx takes over in agent mode. It automatically:
Performs EDA (exploratory data analysis)
Cleans and prepares the dataset (fixing missing values, parsing columns, etc.)
Analyzes numerical and categorical features
Splits the data into training and validation sets
Builds a logistic regression classifier using pipelines, one-hot encoding, and column transformers
Evaluates the model with ROC curves and confusion matrices
Analyzes feature importance to see which factors drive churn

I was impressed by how closely its workflow matched how I would normally approach the problem myself, but it did everything much faster. It even recovered from errors on its own and continued the process without me intervening.

At the end, I also try training an XGBoost model to compare results, showing how flexible and powerful Sphinx can be for real-world machine learning projects.

If you’re into data science, ML, or AI coding assistants, this demo should give you a good sense of what Sphinx can do.

👉 Try it out yourself: https://www.sphinx.ai/

Timecodes:

0:00 Intro – What is Sphinx and setup in VS Code
0:15 Defining the churn classification task
0:44 Sphinx in agent mode: EDA and data cleaning
2:23 Exploring numerical & categorical features
3:41 Preparing data and splitting train/validation
4:03 Building logistic regression with pipelines
5:09 Model evaluation and error handling
6:06 Feature importance and churn insights
7:27 Training XGBoost and comparing models
8:02 Wrap-up and final thoughts