Skip to content

Integrations and deployed versions

Beyond notebooks, pipelines, and serving, Aidin includes the wider Kubeflow ecosystem tooling, deployed and version-pinned by Aidin.

Capability What it does Availability
Katib Hyperparameter search and AutoML (Random, Grid, Bayesian, Hyperband) over your workloads Available
Experiment tracking Automatic tracking of run parameters, status, and metrics via pipelines Available
TensorBoard Training visualization, launched against logs in your workspace storage or a bucket Available
Distributed training operators Multi-GPU / multi-node training jobs (PyTorchJob, TFJob class operators) Available
Workload monitoring GPU/CPU/memory utilization dashboards (Grafana) Available

Aidin docs describe the versions Aidin deploys — upstream documentation for other versions may not match your platform. Current deployed versions (2026-11-10):

Component Deployed version
Notebook controller (Kubeflow Notebooks) v1.9.0
Kubeflow Pipelines (KFP) 2.14.3
KServe 0.20
Katib 0.17

These pins are the compatibility baseline for this documentation set: any instruction copied or adapted from upstream projects is validated against these versions before publication. Where Aidin has modified upstream behavior, the Aidin behavior documented here takes precedence.

Aidin builds on open-source projects, principally the Kubeflow ecosystem (Apache-2.0 licensed code), KServe, and Katib. Upstream names are used for identification. This documentation is written for the Aidin platform; no upstream documentation text is currently reproduced on this site. If adapted upstream text is introduced later, its source, license (Kubeflow website content is licensed CC-BY-4.0), attribution, and modification notice will be recorded on the page that contains it.