The Customer

Société Générale is one of the world's largest financial institutions, headquartered in Paris and operating in more than 60 countries. Its IT Services organization provides private cloud shared infrastructure to every business line in the bank, including application hosting, network services, monitoring and orchestration.

Within IT Services, the Cloud Orchestration Service team is responsible for delivering Apache Airflow as an internal platform service. The team provides a fully managed, on-demand Airflow environment so that engineering teams across the bank can focus on building workflows rather than managing infrastructure. The service is part of Société Générale's broader "Go to Cloud" initiative and the SG Cloud platform, which standardizes infrastructure and enables internal teams to consume cloud-native services on demand.

As Diego Valentierra, Service Manager and Product Owner, explains: "We deliver a standardized, on-demand Airflow platform that abstracts the underlying infrastructure, including Kubernetes, the message broker, and the database. This allows our internal clients to focus on business logic and DAG development rather than platform management."

This model delivers Airflow-as-a-service inside a highly regulated banking environment, supporting a broad range of users across data, security, infrastructure, and application teams.

The Challenge

Before standardizing on Airflow, Société Générale operated multiple legacy orchestration tools across the enterprise. Each system required separate expertise, licensing, and operational overhead, creating fragmentation and slowing modernization efforts. This fragmentation also ran counter to the bank's broader "Go to Cloud" initiative and SG Cloud platform, which aimed to standardize infrastructure and deliver services on demand.

When the team set out to adopt an orchestration tool usable by all IT teams, they evaluated several open-source options as part of a broader push to standardize on open source technologies and selected Apache Airflow. As Aghiles Guezi, DevOps Engineer, recalls, "We compared different open source solutions, and chose Airflow for its flexibility and the strength of its community in 2020."

However, scaling Airflow into a bank-grade platform introduced a new set of challenges:

• Manual provisioning at scale: Early deployments were manually provisioned on virtual machines. Teams requested instances via email, and provisioning could take up to a week, which could not scale to enterprise demand.
• Regulatory and architectural constraints: Strict banking regulations required database isolation and tight infrastructure control, limiting the ability to adopt managed solutions and forcing the team to build and operate the platform internally.
• Expertise gap at scale: Rapid adoption, with hundreds of deployments and a growing number of engineers writing DAGs, created a need for deep Airflow expertise to ensure reliability, performance, and consistent best practices.
• Security labels: With many teams and projects sharing the platform, it became necessary to integrate security labels to enforce data classification policies, control access, ensure isolation between different sensitivity levels, and meet audit and compliance requirements.

The Solution

To accelerate development and ensure production-grade reliability, Société Générale partnered with Astronomer as a strategic Airflow support partner. The engagement provides ongoing technical support, architectural guidance, and access to updates and best practices for operating Airflow at scale within a yearly support contract. Astronomer's award-winning support organization [1] brings deep Airflow expertise and proven customer service excellence to help teams operate with confidence at scale.

With Astronomer's support, Société Générale built the Cloud Orchestration Service, a fully automated, Kubernetes-based Airflow platform that enables teams throughout the enterprise to provision environments on demand.

Astronomer works closely with the platform team through a combination of ticket-based support, advisory services, and regular architectural sessions, helping address complex technical challenges such as:

• Designing scalable Airflow architectures under strict security and isolation requirements (architectural reviews and deployment planning)
• Managing distributed task execution and long-running operations
• Optimizing Kubernetes-based deployments and queueing systems (performance tuning, deployment sizing, and infrastructure optimization)
• Supporting version upgrades, migration strategies, and currently helping upgrade to Airflow 3
• Establishing DAG authoring best practices across a large, distributed user base, including DAG code reviews and authoring guidance

Astronomer also provides enablement and training resources—including workshops and Airflow fundamentals sessions—helping the team build internal expertise and onboard new users as adoption grows. Additional support includes guidance on CI/CD pipeline design, code management strategies, and platform health checks to ensure long-term reliability.

[1] Astronomer was named a 2026 Organization of the Year and awarded the Excellence in Customer Service Award from the Business Intelligence Group.

The Results

With Astronomer support, Société Générale transformed Airflow from a manually provisioned tool into a bank-grade, self-service platform operating at enterprise scale. Key results include:

• From days to on-demand: Provisioning shifted from email-based requests that took up to a week to fully automated, self-service deployments available on demand. "Clients requested Airflow instances by email and we provisioned in about a week… now it's on demand and we currently support over 500 production instances," says Aghiles.
• Enterprise-wide standardization: A single Airflow platform now serves 1,000+ engineers across all business lines, replacing fragmented, team-specific schedulers. As Diego notes, "We provide the platform and all business units use it for their use cases."
• Rapid legacy consolidation: Teams are migrating off other tools —sometimes completing full migrations in a matter of months—reducing licensing costs and operational overhead. According to Diego, "One team evaluated the cost of their legacy scheduler and moved everything to Airflow. In four months, they were completely migrated," says Diego.
• Built-in reliability at scale: Standardized architecture and enforced DAG best practices prevent poorly designed workflows from degrading platform performance, even across hundreds of deployments. "With Astronomer's guidance, we continuously improve DAG design and platform performance at scale," notes Diego.
• Platform leverage for a small team: The Cloud Orchestration team delivers and operates the entire infrastructure layer—Kubernetes, message broker, and database—allowing internal teams to focus purely on business logic. "We remove the infrastructure overhead for our clients—they just write the DAGs," says Diego.
• Faster onboarding for new teams: Airflow's Python-based model makes it easier for a wide range of engineers to adopt the platform quickly, reducing the barrier to entry compared to legacy tools. "Even teams without development experience are quickly able to build their use cases, because Airflow is based on Python and is easier to pick up than legacy tools," says Aghiles.

Airflow is now positioned as a strategic orchestration layer within Société Générale, aligned with the SG Cloud platform and broader enterprise cloud transformation.

What's Next

Demand for the Cloud Orchestration Service platform continues to accelerate as Airflow 3 becomes the default orchestration standard across the bank.

To support this growth, the team is focused on three key priorities:

• Scale and next-gen capabilities: Société Générale is scaling the platform further and unlocking new capabilities with Airflow 3.
• Architectural evolution: The team is evaluating improvements such as enhanced performance and more flexible execution models to support increasingly complex, distributed workloads.
• Migration and enablement at scale: As more teams move off legacy systems, Société Générale is expanding training and enablement programs to accelerate migrations and support users.
• Public cloud extension: On-premises Airflow instances performing orchestration for public cloud resources.

As adoption increases, Astronomer remains a key partner in helping the team navigate new challenges, optimize performance, and continue evolving its Airflow platform at enterprise scale.