Maxis

An AI Training Platform that connects AI model development, training, and evaluation in one place.

Write code in DevSpace, build models with Text Generation, Tabular, and Pipeline Builder, and evaluate and manage results with Benchmark, Model Registry, and Model Compare. Administrators operate training resources with Kueue scheduling and cluster monitoring.

Trained models are served with AI Inference in Metis.

Components

AI Training

Provides training and evaluation environments — DevSpace, Text Generation, Tabular, Pipeline Builder, Benchmark, Model Registry — along with workloads, templates, and volumes.

→

AI Training Admin

View all experiments and workloads from an administrator perspective, and manage Kueue scheduling, monitoring, and projects.

→

Quick Start

If you're new to Maxis, try these recommended guides!

1. Using Pipeline Builder in AI Training

a. Navigate to Pipeline Builder in AI Training.
b. Add nodes for data processing, training, evaluation, and deployment stages.
c. Configure models, datasets, execution environments, and resources for each stage.
d. Save and run the pipeline, then check the execution status and results for each step.

2. Training a Text Generation Model in AI Training

a. Create a Text Generation task in AI Training.
b. Select the base model and training dataset to use.
c. Configure the training method (SFT, DPO, LoRA/QLoRA) and hyperparameters.
d. Run training and review the resulting model and logs.

3. Developing a Tabular Model in AI Training

a. Create a Tabular task in AI Training.
b. Select a structured dataset and specify the target column.
c. Configure task type (prediction, classification) and training settings.
d. Run training and review model performance metrics and results.