1
Team environments
Dataflow gives every teammate the same managed environment instead of ad-hoc notebook-level package installs.
Cloud Data Science Workspace - Jupyter, VS Code and MLflow, Pre-Configured
Start Building Instantly. Skip the setup and get a fully configured workspace with the compute, environments, and apps you need in seconds.
Get Startedin the Dataflow app
Everything You Need, Preconfigured
A complete development workspace designed for data engineers and developers. No configuration required.
Choose Your Compute
Select CPU or GPU instances based on your workload. Scale up for training models, scale down for ETL jobs.
Pre-Installed Applications
VS Code, Jupyter, Airflow, and more—fully configured and ready to use. No manual setup or installation.
Managed Python Environments
Shared, reproducible environments across all applications. No dependency conflicts, no setup time.
Secrets & Credentials
Securely store and access API keys, database passwords, and cloud credentials. Available across all apps automatically.
Database Connections
Connect to PostgreSQL, MySQL, Snowflake, and more. Configured once, available everywhere.
All-in-One Workspace
Compute, environments, secrets, connections, and applications—managed in one unified platform.
From Idea to Execution in Minutes
Traditional setups take hours or days. Install Python, configure virtual environments, set up databases, install applications, manage credentials. Dataflow removes all of that. You go from zero to building in minutes.
Get Startedin the Dataflow app 
Comparison
How Dataflow compares to Google Colab
For Google Colab users, Dataflow keeps the notebook experience but adds persistent team workspaces, shared environments, and production-grade deployment paths. For Databricks users, it is a cost-effective alternative with pre-configured tooling, faster onboarding, and less platform overhead.
2
Persistent workspaces
Your files, apps, and dependencies stay available between sessions so teams do not rebuild context every day.
3
Pipeline integration
Jupyter work connects directly to Airflow orchestration and production deployment without hand-offs.
4
EU data residency
You can deploy on sovereign European cloud providers to satisfy strict governance requirements.
Right-Sized Compute for Every Workload
Choose the instance type that matches your task. Training machine learning models? Spin up a GPU instance. Running scheduled ETL jobs? Use CPU. Switch between them as needed.
CPU Instances
Optimized for data transformations, ETL pipelines, and general-purpose workloads.
GPU Instances
Accelerated compute for ML training, deep learning, and compute-intensive tasks.
On-Demand Scaling
Scale instances up or down based on workload. Pay only for what you use.
To harden your workflow, pair this with Managed Dependencies and Deploy to Production, then compare options on Pricing.