Machine Learning in XGENIA
XGENIA integrates with HuggingFace to bring machine learning capabilities directly into your visual workflows. You can analyze data, train custom models with AutoTrain, and run inference — all without leaving the editor.
What can you do?
- Analyze datasets to detect problem types, suggest targets and features, and assess data quality
- Train models using HuggingFace AutoTrain for text classification, image classification, tabular ML, and LLM fine-tuning
- Run inference with any HuggingFace model — serverless or via dedicated endpoints
- Predict customer retention using AI-powered churn analysis
- Scaffold ML pipelines on the canvas with a single AI command
Architecture
The ML system has three layers:
| Layer | Description |
|---|---|
| AI Tools | The XGENIA AI assistant has built-in ML tools (hf_analyze_data, hf_predict, autotrain_create_project, etc.) that you can invoke through chat |
| ML Coordinator Server | A lightweight Node.js server that wraps HuggingFace APIs (AutoTrain, Inference) and provides analysis endpoints |
| Pro Nodes | Visual nodes you can drag onto the canvas for data analysis, model training, prediction, and retention analysis |