Choosing a sandbox runtime for long-lived coding agents
How we evaluated sandbox providers for coding agents, what we chose for now, and the trade-offs behind that decision.
Start locally without complicated setup hassle
- ZenML is available as a simple pip package that lets you run and track pipelines locally.
- ZenML integrates with your orchestration layer of choice, avoiding having to learn different paradigms for dev, staging, and prod.
- ZenML integrates with your orchestration layer of choice or can be extended with your own orchestration service.
Abstract away infrastructure complexity
- Most orchestrators assume some form of infrastructure knowledge to use them maximally — ZenML abstracts that complexity away.
- ZenML separates infrastructure setup like Docker building from the application logic, and automates the tedious parts.
- ZenML focuses on the handovers between MLOps Engineers, ML Engineers, and Data Scientists.
Switch between orchestrators depending on your context
- You can switch between different orchestration services with a single click — from dev to staging to production.
- The more engineering-minded in the team still retain control over their productionalization because the framework is extensible.
- ZenML handles the pain of packaging your code into Docker to be deployed to your orchestration service of choice.
After a benchmark on several solutions, we choose ZenML for its stack flexibility and its incremental process. We started from small local pipelines and gradually created more complex production ones. It was very easy to adopt.
Clément Depraz
Data Scientist at Brevo
Explore in Detail What Makes ZenML Unique
| Feature | ZenML | Kubeflow |
|---|---|---|
| Workflow Orchestration | Yes Provides a flexible and user-friendly orchestration layer on top of Kubeflow | Yes Offers powerful Kubernetes-native orchestration for ML workflows |
| Ease of Use | Yes Simplifies the adoption and management of Kubeflow pipelines with an intuitive interface | Not supported Requires Kubernetes expertise to effectively utilize its features |
| Integration Flexibility | Yes Seamlessly integrates Kubeflow with other MLOps tools for a customized stack | Not supported Primarily focuses on Kubernetes-based integrations and extensions |
| Switch your orchestrator, keep your code | Yes Keep the same code when switching orchestrator | Not supported Requires significant rewriting to use Kubeflow code with a different orchestrator |
| Pipeline Customization | Yes Enables easy customization and extension of Kubeflow pipelines | Yes Allows customization but may require more Kubernetes knowledge |
| Collaborative MLOps | Yes Facilitates collaboration among teams with version control and governance features | Yes Provides collaboration features but may require additional setup |
| Scalability | Yes Leverages Kubeflow's scalability while providing an abstraction layer for ease of use | Yes Highly scalable for large-scale ML workflows on Kubernetes |
| Experiment Tracking | Yes Integrates with MLflow and other tools for comprehensive experiment tracking | Yes Offers Kubeflow Metadata for experiment tracking and artifact management |
| Model Deployment | Yes Simplifies the deployment of models using Kubeflow with pre-built integrations | Yes Supports various deployment options, including Kubeflow Serving |
| Monitoring and Logging | Yes Provides centralized monitoring and logging for Kubeflow pipelines | Yes Offers Kubeflow Metadata for logging and monitoring |
| Community and Support | Yes Growing community with active support and resources | Yes Large and active community with extensive resources and support |
| MLOps Lifecycle Coverage | Yes Covers the entire MLOps lifecycle, from data preparation to model monitoring | Not supported Focuses primarily on orchestration, deployment, and serving |
| Learning Curve | Yes Reduces the learning curve for adopting Kubeflow with a user-friendly abstraction layer | Not supported Requires Kubernetes expertise to effectively utilize its full set of features |
| Hybrid and Multi-Cloud | Yes Supports hybrid and multi-cloud deployments with Kubeflow integration | Yes Enables hybrid and multi-cloud deployments on Kubernetes |
| GPU and Distributed Computing | Yes Seamlessly leverages Kubeflow's GPU and distributed computing capabilities | Yes Provides strong support for GPU and distributed computing workloads |
Code comparison
ZenML and Kubeflow side by side
# zenml integration install kubeflow
# zenml orchestrator register kf_orchestrator -f kubeflow ...
# zenml stack update my_stack -o kf_orchestrator
from zenml import pipeline, step
@step
def preprocess_data(data_path: str) -> str:
# Preprocessing logic here
return processed_data
@step
def train_model(data: str):
# Model training logic here
return model
@pipeline
def my_pipeline(data_path: str):
processed_data = preprocess_data(data_path)
model = train_model(processed_data)
# Run the pipeline
my_pipeline(data_path="path/to/data")# Kubeflow Pipelines SDK
import kfp
from kfp import dsl
def preprocess_op(data_path):
return dsl.ContainerOp(
name='Preprocess Data',
image='preprocess-image:latest',
arguments=['--data_path', data_path]
)
def train_op(data):
return dsl.ContainerOp(
name='Train Model',
image='train-image:latest',
arguments=['--data', data]
)
@dsl.pipeline(
name='My ML Pipeline',
description='A sample ML pipeline'
)
def my_pipeline(data_path: str):
preprocess_task = preprocess_op(data_path)
train_task = train_op(preprocess_task.output)
# Compile and run the pipeline
kfp.compiler.Compiler().compile(my_pipeline, 'pipeline.yaml')
client = kfp.Client()
client.create_run_from_pipeline_func(my_pipeline, arguments={})01.
Streamlined ML Workflow Initialization
ZenML guarantees swifter initialization, surpassing orchestrators for prompt, optimized ML workflows.
02.
Supporting All Your Tools
ZenML is a native interface to the whole end-to-end machine learning lifecycle, taking you beyond just orchestration.
03.
Unrivaled User Assistance
ZenML excels with dedicated support, offering personalized assistance beyond standard orchestrators.
Outperform Orchestrators: Book Your Free ZenML Strategy Talk
Orchestrator Showdown
Explore the Advantages of ZenML Over Other Orchestrator Tools
ZenML vs Apache Airflow
ML-optimized workflow management. Enhance scalability and usability with comprehensive features designed for ML pipeline orchestration.
ZenML vs Argo Workflows
ZenML is an open-source alternative to Argo Workflows for ML pipelines with built-in metadata, lineage, and reproducibility
ZenML vs Dagster
Dagster alternative: Streamline ML ops with intuitive pipelines. Seamless integrations and experiment tracking for efficient ML workflow management.
ZenML vs Databricks
Databricks alternative: Flexible ML orchestration without vendor lock-in. Accelerate ML with lightweight, adaptable workflows across multiple clouds.
ZenML vs Flyte
Flyte alternative: Agile ML pipeline orchestration. Accelerate workflows with intuitive tools, seamless MLOps integration, and rapid iteration.
ZenML vs Kedro
Kedro alternative: Scalable, user-friendly ML framework. Streamline operations with robust features for efficient project management and deployment.
ZenML vs Prefect
Prefect alternative: ML-centric pipeline orchestration. Streamline workflows with intuitive design, experiment tracking, and MLOps integrations.
Broaden Your AI Orchestration Understanding with ZenML
Introducing the new Kitaru: from production traces to repeatable evals
The new Kitaru turns production traces you already collect into replayable test scenarios, expert-reviewed cohorts, and evaluators you can run on every change.
Your GPUs Are Everywhere. Your Robot-Learning Loop Shouldn't Be.
Robotics compute is spreading across clouds and clusters. Learn how one portable pipeline layer can keep the robot-learning loop reproducible.
Experience the ZenML Difference:
Book Your Customized Demo
See ZenML's superior model orchestration in action Discover how ZenML offers more with your existing ML tools Find out why data security with ZenML outshines the rest