The Data Flowcast: Mastering Apache Airflow ® for Data Engineering and AI

Latest Episodes:

The Software Risk That Affects Everyone and How To Address It with Michael Winser and Jarek Potiuk Mar 20, 2025

Show notes

The security of open-source software is a growing concern, especially as dependencies and regulations become more complex, making it essential to understand how to manage software supply chains effectively.

In this episode, we sit down with Michael Winser, Co-Founder at Alpha-Omega and Security Strategy Ambassador at Eclipse Foundation, and Jarek Potiuk, Member of the Security Committee at the Apache Software Foundation, to discuss the challenges of securing Airflow’s dependencies, the evolving landscape of open-source security and how contributors can help strengthen the ecosystem.

Key Takeaways:

(02:43) Jarek quit his full-time engineer position and uses Airflow as a freelancer.

(04:32) Michael finds happiness in having meaningful work with open-source security.

(07:01) Software supply chain security focuses on correctness, integrity and availability.

(08:44) Airflow’s 790 dependencies present a unique security challenge.

(09:43) Airflow’s security team has significantly improved its vulnerability response.

(10:22) The transition to Airflow 3 emphasizes enterprise security readiness.

(16:20) The ‘Three Fs’ approach: fix it, fork it, or forget it.

(18:45) Dependency health is often more critical than fixing known vulnerabilities.

(23:32) The ‘Three Fs’ in action.

(26:26) Open-source contributors play a key role in supply chain security.

Resources Mentioned:

Michael Winser -

https://www.linkedin.com/in/michaelw/

Jarek Potiuk -

https://www.linkedin.com/in/jarekpotiuk/

Apache Airflow -

https://airflow.apache.org/

Apache Software Foundation | LinkedIn -

https://www.linkedin.com/company/the-apache-software-foundation/

Apache Software Foundation | Website -

https://www.apache.org/

Eclipse Foundation | LinkedIn -

https://www.linkedin.com/company/eclipse-foundation/

Eclipse Foundation | Website -

https://www.eclipse.org/org/foundation/

OpenSSF Working Groups -

https://openssf.org/community/openssf-working-groups/

Astronomer Roadshow: Exploring Apache Airflow 3 | London

https://www.astronomer.io/events/roadshow/london/

Astronomer Roadshow: Exploring Apache Airflow 3 | New York

https://www.astronomer.io/events/roadshow/new-york/

Astronomer Roadshow: Exploring Apache Airflow 3 | Sydney

https://www.astronomer.io/events/roadshow/sydney/

Astronomer Roadshow: Exploring Apache Airflow 3 | San Francisco

https://www.astronomer.io/events/roadshow/san-francisco/

Astronomer Roadshow: Exploring Apache Airflow 3 | Chicago

https://www.astronomer.io/events/roadshow/chicago/

Thanks for listening to “The Data Flowcast: Mastering Airflow for Data Engineering & AI.” If you enjoyed this episode, please leave a 5-star review to help get the word out about the show. And be sure to subscribe so you never miss any of the insightful conversations.

#AI #Automation #Airflow #MachineLearning

Building Scalable ML Infrastructure at Outerbounds with Savin Goyal Mar 13, 2025

Show notes

Machine learning is changing fast, and companies need better tools to handle AI workloads. The right infrastructure helps data scientists focus on solving problems instead of managing complex systems. In this episode, we talk with Savin Goyal, Co-Founder and CTO at Outerbounds, about building ML infrastructure, how orchestration makes workflows easier and how Metaflow and Airflow work together to simplify data science.

Key Takeaways:

(02:02) Savin spent years building AI and ML infrastructure, including at Netflix.

(04:05) ML engineering was not a defined role a decade ago.

(08:17) Modernizing AI and ML requires balancing new tools with existing strengths.

(10:28) ML workloads can be long-running or require heavy computation.

(15:29) Different teams at Netflix used multiple orchestration systems for specific needs.

(20:10) Stable APIs prevent rework and keep projects moving.

(21:07) Metaflow simplifies ML workflows by optimizing data and compute interactions.

(25:53) Limited local computing power makes running ML workloads challenging.

(27:43) Airflow UI monitors pipelines, while Metaflow UI gives ML insights.

(33:13) The most successful data professionals focus on business impact, not just technology.

Resources Mentioned:

Savin Goyal -

https://www.linkedin.com/in/savingoyal/

Outerbounds -

https://www.linkedin.com/company/outerbounds/

Apache Airflow -

https://airflow.apache.org/

Metaflow -

https://metaflow.org/

Netflix’s Maestro Orchestration System -

https://netflixtechblog.com/maestro-netflixs-workflow-orchestrator-ee13a06f9c78?gi=8e6a067a92e9#:~:text=Maestro%20is%20a%20fully%20managed,data%20between%20different%20storages%2C%20etc.

TensorFlow -

https://www.tensorflow.org/

PyTorch -

https://pytorch.org/

#AI #Automation #Airflow #MachineLearning

Customizing Airflow for Complex Data Environments at Stripe with Nick Bilozerov and Sharadh Krishnamurthy Mar 06, 2025

Show notes

Keeping data pipelines reliable at scale requires more than just the right tools — it demands constant innovation. In this episode, Nick Bilozerov, Senior Data Engineer at Stripe, and Sharadh Krishnamurthy, Engineering Manager at Stripe, discuss how Stripe customizes Airflow for its needs, the evolution of its data orchestration framework and the transition to Airflow 2. They also share insights on scaling data workflows while maintaining performance, reliability and developer experience.

Key Takeaways:

(02:04) Stripe’s mission is to grow the GDP of the internet by supporting businesses with payments and data.

(05:08) 80% of Stripe engineers use data orchestration, making scalability critical.

(06:06) Airflow powers business reports, regulatory needs and ML workflows.

(08:02) Custom task frameworks improve dependencies and validation.

(08:50) "User scope mode" enables local testing without production impact.

(10:39) Migrating to Airflow 2 improves isolation, safety and scalability.

(16:40) Monolithic DAGs caused database issues, prompting a service-based shift.

(19:24) Frequent Airflow upgrades ensure stability and access to new features.

(21:38) DAG versioning and backfill improvements enhance developer experience.

(23:38) Greater UI customization would offer more flexibility.

Resources Mentioned:

Nick Bilozerov -

https://www.linkedin.com/in/nick-bilozerov/

Sharadh Krishnamurthy -

https://www.linkedin.com/in/sharadhk/

Apache Airflow -

https://airflow.apache.org/

Stripe | LinkedIn -

https://www.linkedin.com/company/stripe/

Stripe | Website -

https://stripe.com/

#AI #Automation #Airflow #MachineLearning

Harnessing Airflow for Data-Driven Policy Research at CSET with Jennifer Melot Feb 27, 2025

Show notes

Turning complex datasets into meaningful analysis requires robust data infrastructure and seamless orchestration. In this episode, we’re joined by Jennifer Melot, Technical Lead at the Center for Security and Emerging Technology (CSET) at Georgetown University, to explore how Airflow powers data-driven insights in technology policy research. Jennifer shares how her team automates workflows to support analysts in navigating complex datasets.

Key Takeaways:

(02:04) CSET provides data-driven analysis to inform government decision-makers.

(03:54) ETL pipelines merge multiple data sources for more comprehensive insights.

(04:20) Airflow is central to automating and streamlining large-scale data ingestion.

(05:11) Larger-scale databases create challenges that require scalable solutions.

(07:20) Dynamic DAG generation simplifies Airflow adoption for non-engineers.

(12:13) DAG Factory and dynamic task mapping can improve workflow efficiency.

(15:46) Tracking data lineage helps teams understand dependencies across DAGs.

(16:14) New Airflow features enhance visibility and debugging for complex pipelines.

Resources Mentioned:

Jennifer Melot -

https://www.linkedin.com/in/jennifer-melot-aa710144/

Center for Security and Emerging Technology (CSET) -

https://www.linkedin.com/company/georgetown-cset/

Apache Airflow -

https://airflow.apache.org/

Zenodo -

https://zenodo.org/

OpenLineage -

https://openlineage.io/

Cloud Dataplex -

https://cloud.google.com/dataplex

#AI #Automation #Airflow #MachineLearning

Hybrid Testing Solutions for Autonomous Driving at Bosch with Jens Scheffler and Christian Schilling Feb 13, 2025

Show notes

Testing autonomous vehicles demands precision, scalability and powerful orchestration tools — enter Apache Airflow, a key component of Bosch’s cutting-edge testing framework. In this episode, we sit down with Jens Scheffler, Test Execution Cluster Technical Architect, and Christian Schilling, Product Owner Open Loop Testing Automated Driving, both at Bosch, to explore how Bosch harnesses Airflow to streamline complex testing scenarios. They share insights on scaling workflows, integrating hybrid infrastructures and ensuring vehicle safety through rigorous automated testing.

Key Takeaways:

(01:35) Airflow orchestrates millions of test hours for autonomous systems.

(03:15) Jens scales distributed systems with Kubernetes for job orchestration.

(06:02) Airflow runs hundreds of tests simultaneously.

(06:44) Virtual testing reduces costs and on-road trials.

(12:19) Unified APIs and GUIs streamline operations.

(15:05) Self-service setups empower Bosch teams.

(18:00) Physical hardware integration ensures real-world timing.

(20:30) Dynamic task mapping scales workflows efficiently.

(25:22) Open-source contributions improve stability.

(31:06) Edge and Celery executors power Bosch's hybrid scheduling.

Resources Mentioned:

Jens Scheffler -

https://www.linkedin.com/in/jens-scheffler/

Christian Schilling -

https://www.linkedin.com/in/christian-schilling-a5078831a/

Bosch -

https://www.linkedin.com/company/bosch/

Apache Airflow -

https://airflow.apache.org/

Kubernetes -

https://kubernetes.io

GitHub -

https://github.com

Edge Executor -

https://airflow.apache.org/docs/apache-airflow/stable/core-concepts/executor/index.html

#AI #Automation #Airflow #MachineLearning

Overcoming Airflow Scaling Challenges at Monzo Bank with Jonathan Rainer Feb 06, 2025

Show notes

Scaling a data orchestration platform to manage thousands of tasks daily demands innovative solutions and strategic problem-solving. In this episode, we explore the complexities of scaling Airflow and the challenges of orchestrating thousands of tasks in dynamic data environments. Jonathan Rainer, Former Platform Engineer at Monzo Bank, joins us to share his journey optimizing data pipelines, overcoming UI limitations and ensuring DAG consistency in high-stakes scenarios.

Key Takeaways:

(03:11) Using Airflow to schedule computation in BigQuery.

(07:02) How DAGs with 8,000+ tasks were managed nightly.

(08:18) Ensuring accuracy in regulatory reporting for banking.

(11:35) Handling task inconsistency and DAG failures with automation.

(16:09) Building a service to resolve DAG consistency issues in Airflow.

(25:05) Challenges with scaling the Airflow UI for thousands of tasks.

(27:03) The role of upstream and downstream task management in Airflow.

(37:33) The importance of operational metrics for monitoring Airflow health.

(39:19) Balancing new tools with root cause analysis to address scaling issues.

(41:35) Why scaling solutions require both technical and leadership buy-in

Resources Mentioned:

Jonathan Rainer -

https://www.linkedin.com/in/jonathan-rainer/

Monzo Bank -

https://www.linkedin.com/company/monzo-bank/

Apache Airflow -

https://airflow.apache.org/

BigQuery -

https://airflow.apache.org/docs/apache-airflow-providers-google/stable/operators/cloud/bigquery.html

Kubernetes -

https://kubernetes.io/

#AI #Automation #Airflow #MachineLearning

Orchestrating Analytics and AI Workflows at Telia with Arjun Anandkumar Jan 30, 2025

Show notes

The future of data engineering lies in seamless orchestration and automation. In this episode, Arjun Anandkumar, Data Engineer at Telia, shares how his team uses Airflow to drive analytics and AI workflows. He highlights the challenges of scaling data platforms and how adopting best practices can simplify complex processes for teams across the organization. Arjun also discusses the transformative role of tools like Cosmos and Terraform in enhancing efficiency and collaboration.

Key Takeaways:

(02:16) Telia operates across the Nordics and Baltics, focusing on telecom and energy services.

(03:45) Airflow runs dbt models seamlessly with Cosmos on AWS MWAA.

(05:47) Cosmos improves visibility and orchestration in Airflow.

(07:00) Medallion Architecture organizes data into bronze, silver and gold layers.

(08:34) Task group challenges highlight the need for adaptable workflows.

(15:04) Scaling managed services requires trial, error and tailored tweaks.

(19:46) Terraform scales infrastructure, while YAML templates manage DAGs efficiently.

(20:00) Templated DAGs and robust testing enhance platform management.

(24:15) Open-source resources drive innovation in Airflow practices.

Resources Mentioned:

Arjun Anandkumar -

https://www.linkedin.com/in/arjunanand1/?originalSubdomain=dk

Telia -

https://www.linkedin.com/company/teliacompany/

Apache Airflow -

https://airflow.apache.org/

Cosmos by Astronomer -

https://www.astronomer.io/cosmos/

Terraform -

https://www.terraform.io/

Medallion Architecture by Databricks -

https://www.databricks.com/glossary/medallion-architecture

#AI #Automation #Airflow #MachineLearning

The Role of Airflow in Finance Transformation at Etraveli Group with Mihir Samant Jan 23, 2025

Show notes

Transforming bottlenecked finance processes into streamlined, automated systems requires the right tools and a forward-thinking approach. In this episode, Mihir Samant, Senior Data Analyst at Etraveli Group, joins us to share how his team leverages Airflow to revolutionize finance automation. With extensive experience in data workflows and a passion for open-source tools, Mihir provides valuable insights into building efficient, scalable systems. We explore the transformative power of Airflow in automating workflows and enhancing data orchestration within the finance domain.

Key Takeaways:

(02:14) Etraveli Group specializes in selling affordable flight tickets and ancillary services.

(03:56) Mihir’s finance automation team uses Airflow to tackle month-end bottlenecks.

(06:00) Airflow's flexibility enables end-to-end automation for finance workflows.

(07:00) Open-source Airflow tools offer cost-effective solutions for new teams.

(08:46) Sensors and dynamic DAGs are pivotal features for optimizing tasks.

(13:30) GitSync simplifies development by syncing environments seamlessly.

(16:27) Plans include integrating Databricks for more advanced data handling.

(17:58) Airflow and Databricks offer multiple flexible methods to trigger workflows and execute SQL queries seamlessly.

Resources Mentioned:

Mihir Samant -

https://www.linkedin.com/in/misamant/?originalSubdomain=ca

Etraveli Group -

https://www.linkedin.com/company/etraveli-group/

Apache Airflow -

https://airflow.apache.org/

Docker -

https://www.docker.com/

Databricks -

https://www.databricks.com/

#AI #Automation #Airflow #MachineLearning

Inside Ford’s Data Transformation: Advanced Orchestration Strategies with Vasantha Kosuri-Marshall Jan 16, 2025

Show notes

Data engineering is entering a new era, where orchestration and automation are redefining how large-scale projects operate. This episode features Vasantha Kosuri-Marshall, Data and ML Ops Engineer at Ford Motor Company. Vasantha shares her expertise in managing complex data pipelines. She takes us through Ford's transition to cloud platforms, the adoption of Airflow and the intricate challenges of orchestrating data in a diverse environment.

Key Takeaways:

(03:10) Vasantha’s transition to the Advanced Driving Assist Systems team at Ford.

(05:42) Early adoption of Airflow to orchestrate complex data pipelines.

(09:29) Ford's move from on-premise data solutions to Google Cloud Platform.

(12:03) The importance of Airflow's scheduling capabilities for efficient data management.

(16:12) Using Kubernetes to scale Airflow for large-scale data processing.

(19:59) Vasantha’s experience in overcoming challenges with legacy orchestration tools.

(22:22) Integration of data engineering and data science pipelines at Ford.

(28:03) How deferrable operators in Airflow improve performance and save costs.

(32:12) Vasantha’s insights into tuning Airflow properties for thousands of DAGs.

(36:09) The significance of monitoring and observability in managing Airflow instances.

Resources Mentioned:

Vasantha Kosuri-Marshall -

https://www.linkedin.com/in/vasantha-kosuri-marshall-0b0aab188/

Apache Airflow -

https://airflow.apache.org/

Google Cloud Platform (GCP) -

https://cloud.google.com/

Ford Motor Company | LinkedIn -

https://www.linkedin.com/company/ford-motor-company/

Ford Motor Company | Website -

https://www.ford.com/

Astronomer -

https://www.astronomer.io/

#AI #Automation #Airflow #MachineLearning

Powering Finance With Advanced Data Solutions at Ramp with Ryan Delgado Jan 10, 2025

Show notes

Data is the backbone of every modern business, but unlocking its full potential requires the right tools and strategies. In this episode, Ryan Delgado, Director of Engineering at Ramp, joins us to explore how innovative data platforms can transform business operations and fuel growth. He shares insights on integrating Apache Airflow, optimizing data workflows and leveraging analytics to enhance customer experiences.

Key Takeaways:

(01:52) Data is the lifeblood of Ramp, touching every vertical in the company.

(03:18) Ramp’s data platform team enables high-velocity scaling through tailored tools.

(05:27) Airflow powers Ramp’s enterprise data warehouse integrations for advanced analytics.

(07:55) Centralizing data in Snowflake simplifies storage and analytics pipelines.

(12:08) Machine learning models at Ramp integrate seamlessly with Airflow for operational excellence.

(14:11) Leveraging Airflow datasets eliminates inefficiencies in DAG dependencies.

(17:22) Platforms evolve from solving narrow business problems to scaling organizationally.

(18:55) ClickHouse enhances Ramp’s OLAP capabilities with 100x performance improvements.

(19:47) Ramp’s OLAP platform improves performance by reducing joins and leveraging ClickHouse.

(21:46) Ryan envisions a lighter-weight, more Python-native future for Airflow.

Resources Mentioned:

Ryan Delgado -

https://www.linkedin.com/in/ryan-delgado-69544568/

Ramp -

https://www.linkedin.com/company/ramp/

Apache Airflow -

https://airflow.apache.org/

Snowflake -

https://www.snowflake.com/

ClickHouse -

https://clickhouse.com/

dbt -

https://www.getdbt.com/

#AI #Automation #Airflow #MachineLearning

Our TOPPODCAST Picks

Follow Us

Stay Connected

The Data Flowcast: Mastering Apache Airflow ® for Data Engineering and AI

Related Podcasts

Links

Stay Connected