watsonx.ai
Solution report card
| Runs on IBM i? | ❌ | |
| On-prem | ✅ | |
| IBM Cloud | ✅ | |
| AI capabilities | Machine Learning AutoAI Deep Learning Generative AI many more… | |
| Commercial support | ✅ | |
| Free to try? | ✅ | |
| Requirements |
What is watsonx.ai?
Section titled “What is watsonx.ai?”watsonx.ai is IBM’s AI development platform — a managed environment for building, training, tuning, and deploying AI models. It serves as IBM’s primary platform for both classical machine learning and generative AI, and is available as a fully managed service on IBM Cloud or deployed on-premises via IBM Cloud Pak for Data.
For IBM i practitioners, watsonx.ai is a natural destination for AI workloads that need more compute than IBM i itself provides, while keeping data within IBM’s enterprise ecosystem.
Key capabilities
Section titled “Key capabilities”Foundation models and generative AI
Section titled “Foundation models and generative AI”watsonx.ai hosts IBM’s Granite family of foundation models — purpose-built enterprise LLMs optimized for business tasks like summarization, classification, code generation, and question answering. Granite models are designed with enterprise governance in mind: IBM provides transparency about training data and indemnification for production use.
Beyond Granite, watsonx.ai provides access to a curated set of open-source foundation models including Llama 3, Mistral, and others.
Prompt Lab
Section titled “Prompt Lab”An interactive environment for experimenting with LLM prompts without writing code. Useful for quickly prototyping natural language interfaces over IBM i data before building a full application.
AutoAI
Section titled “AutoAI”AutoAI automates the machine learning pipeline — feature engineering, algorithm selection, hyperparameter tuning — allowing users without deep ML expertise to build high-quality models from tabular data. IBM i’s Db2 for i data can be imported directly as training input.
Model training and tuning
Section titled “Model training and tuning”watsonx.ai supports fine-tuning foundation models on domain-specific data, full model training with notebooks (Jupyter-based), and integration with popular frameworks (scikit-learn, PyTorch, XGBoost).
Deployment and serving
Section titled “Deployment and serving”Trained models deploy to a managed inference endpoint with a REST API, making it straightforward to call from IBM i applications using SQL HTTP functions or the Db2 for i AI SDK.
Connecting watsonx.ai to IBM i data
Section titled “Connecting watsonx.ai to IBM i data”See Accessing Db2 from watsonx.ai for step-by-step instructions on creating a Db2 for i connection in watsonx.ai projects.