Sematic

Sematic is an open-source ML orchestration platform built for teams that want to define, run, and track training pipelines in pure Python rather than wrestling with YAML, Kubeflow CRDs, or a homegrown Airflow fork. You install it with `pip install sematic`, decorate functions, and the same pipeline that runs on your laptop can be executed on a Kubernetes cluster with automatic environment packaging, artifact versioning, and a dashboard for visualizing DAGs, inputs, outputs, and reruns.

It targets ML engineers and platform teams who have outgrown ad-hoc notebooks but don't want the operational weight of a full ML platform like Kubeflow or Flyte. The differentiator is the Python-native, declarative API with type-checked inputs/outputs and nested/dynamic graphs, plus a UI that treats every pipeline run as a first-class, inspectable artifact. The core is free and open source under Apache 2.0; a hosted/enterprise tier exists for teams that want managed infrastructure and support.

It integrates with the usual ML stack (PyTorch, HuggingFace, Ray, Snowflake, S3) and runs on Kubernetes for cloud execution. Note that Sematic is orchestration infrastructure for ML, not an LLM or generative tool itself, and the open-source project's release cadence has been quieter recently than competitors like Prefect or Dagster.