Getting started with AION
1. Accessing AION
- Cloud Deployment: Log in to the AION web platform.
- On-Prem Installation: Install AION using Docker/Kubernetes based on your environment.
- API Access: Use AION’s API suite for seamless integration with your existing system.
2. Upload and prepare data
- Data Ingestion: Upload CSV, JSON, or connect to databases (PostgreSQL, Actian, etc.).
- EDA & Feature Engineering: Analyzes data, detects missing values, and suggests transformations.
3. Build and train models
- AutoML: Select the best model automatically.
- Custom Model: Create the best model based on a vast set of problem types and algorithms.
- Bring Your Own Model (BYOM): Upload and deploy custom ML models (TensorFlow, PyTorch, Scikit-Learn).
- MLOps Workflow: Automate model retraining based on drift detection.
4. Deploy and monitor models
Deployment Options:
- REST API: HTTP endpoints for inference.
- Model Stubs: Lightweight SDK for easy integration.
- Containerised Models: Docker-based deployment.
Performance Monitoring:
- Track latency, accuracy, and drift in real-time.
5. Explain and optimize AI designs
- Model Explainability: SHAP/RFE-based feature importance.
- Drift Detection: Detect and flag changes in data patterns.
- AI Governance: Maintain model lineage, versioning, and logs.