The Customer

BearingPoint is a leading global consultancy known for delivering high-impact technology and data-driven solutions to enterprise organizations. Its Products division operates like a SaaS company, building, owning, and delivering data and AI applications directly to enterprise customers.

Led by Julien Damon, CTO of the Products division in France, the team develops production-grade ML-driven data products, including retail demand forecasting and credit risk platforms. In these environments, data pipelines are not just infrastructure. They are mission-critical systems that power the data products customers rely on every day.

As Julien puts it, "Our products are built on data. If the pipelines are not reliable, the product is not reliable."

The Challenge

Before Airflow, BearingPoint relied on a fully custom Python orchestration framework. The team effectively built core orchestration capabilities themselves, including scheduling, dependency management, and monitoring from scratch.

What worked initially became difficult to scale and maintain as the platform grew:

• Lack of visibility into failures: Issues were sometimes first noticed by enterprise customers rather than detected proactively by the team
• ~10 incidents per week: Frequent pipeline failures disrupted production workflows
• High operational burden: Engineers spent too much time keeping the lights on and performing debugging work, and needed to spend more time on new product development
• Multi-cloud requirements: The team needed an orchestration solution that could operate consistently across AWS and Azure

With only a handful of engineers per SaaS product, every failure had a direct cost. Debugging pipelines meant delayed features, slower iteration, and increased risk to customer SLAs.

"As our platform grew, we saw the need for stronger visibility and reliability in our data pipelines to support the products our customers depend on." Julien Damon CTO, Products France, BearingPoint

The Solution

BearingPoint standardized on Apache Airflow to adopt an open, proven orchestration standard that their engineers could extend and rely on across use cases. As Julien explains:

"We chose Airflow because it gives us an industry standard way to define, schedule, and monitor pipelines in code, with the flexibility to support both data, ML, and emerging AI workloads." Julien Damon CTO, Products France, BearingPoint

The team then needed a way to run Airflow that could support a multi-cloud environment, operate with a lean team, and meet production-grade reliability requirements.

After evaluating alternatives, the team chose Astronomer for its cloud-agnostic approach and its deep investment in the Airflow ecosystem.

“Hyperscaler solutions are not cloud-agnostic. With Astronomer and Astro, we have a foundation that works consistently across AWS, Azure, and other providers, and is sustainable, supported, and reliable for the production systems our customers depend on every day." Julien Damon CTO, Products France, BearingPoint

Today, Astro operates as the orchestration backbone for BearingPoint’s SaaS platform. The following examples highlight how Astro supports production applications:

Credit Scoring

BearingPoint’s Sellia platform powers ML-driven credit risk decisions for enterprise financial services customers, including B2B lenders. Astro orchestrates reliable, production-grade batch pipelines that support data preparation, model training, and credit scoring.

Astro coordinates workflows that ingest data from financial systems, validate and transform datasets, generate features, retrain or update models, and produce batch credit scores delivered to downstream risk and decision systems. These pipelines are designed for consistency, traceability, and governance, which are critical in regulated financial environments.

Reusable pipeline steps handle data validation, feature generation, and model inference, ensuring a consistent and observable pattern across all scoring workflows.

Retail Forecasting

BearingPoint’s DemandSens platform delivers ML-driven retail forecasting, helping enterprise retailers optimize inventory and demand planning. Astro orchestrates both nightly batch workflows and real-time data pipelines to support these use cases.

Each night, Astro coordinates an end-to-end workflow that ingests point-of-sale data, validates and transforms datasets, performs feature engineering, runs ML models, and publishes next-day forecasts to downstream applications. In parallel, Astro orchestrates real-time updates using streaming or incremental data to refresh features and predictions throughout the day.

DAGs enforce dependencies and SLAs across both workflows, with alerts on delays to ensure forecasts remain accurate and available when needed for operational decision-making.

Across both applications, Astro provides a consistent operational model. It reduces complexity while enabling flexibility across clouds.

The Results

Astro transformed orchestration from a bottleneck into a scalable foundation for growth:

• 90% reduction in pipeline incidents, driven by standardized DAG patterns, retries, and dependency management that reduce failure rates
• 95% faster MTTR with centralized logs, alerting, and task-level visibility for faster root cause analysis
• 6% engineering time savings, enabled by built-in CI/CD, Git-based workflows, automated testing, and one-click deployments that reduce manual effort and accelerate iteration
• 95%+ SLA attainment, supported by scheduling, SLA monitoring, and proactive alerting on delays
• Team efficiency: 0 additional hires required to manage orchestration because Astro simplified operations while Astronomer offers world-class support

These results are driven by Astro’s core capabilities, including cloud-agnostic orchestration across AWS and Azure, managed Airflow that removes infrastructure overhead, and built-in observability and CI/CD that accelerate debugging and deployment.

"Astro gives us the reliability and visibility we need to run critical data, ML, and AI products at scale. We can focus on delivering value to our customers instead of managing pipelines." Julien Damon CTO, Products France, BearingPoint

What’s Next

BearingPoint is continuing to migrate remaining applications off its legacy framework and standardizing on Airflow and Astro across its entire product portfolio.

Looking ahead, the team is expanding into new European markets and investing in AI-native applications, including products that move beyond traditional ETL patterns.

With a unified orchestration layer in place, BearingPoint can scale new use cases faster without rethinking its infrastructure each time.

"We don’t want to rely on homemade frameworks. With Astronomer and Astro, we have an orchestration foundation that is sustainable, supported, and reliable for running production systems our customers depend on every day." Julien Damon CTO, Products France, BearingPoint