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

Latest Episodes:

Introducing Airflow’s Common AI Provider with Pavan Kumar Gopidesu and Kaxil Naik Apr 23, 2026

Show notes

In this episode, we explore the newly released Apache Airflow common AI provider — what problem it solves, how it was built and what's coming next.

Kaxil Naik, Senior Director of Engineering at Astronomer and Apache Airflow PMC member, and Pavan Kumar Gopidesu, Lead Data Engineer at Experian and Apache Airflow PMC member, join us to walk through the provider's first release and the technical decisions behind it.

Key Takeaways:

00:00 Introduction.

04:05 The common AI provider was born from a real production problem.

07:10 Airflow already had the primitives needed for durable agent execution, making it the natural foundation for AI orchestration.

09:15 The LLM schema compare operator uses Apache DataFusion to fetch source schemas.

11:07 Apache DataFusion was chosen for its speed.

13:09 Hook tool sets expose Airflow's provider hooks to agents with an allowed methods list that blocks destructive operations.

15:20 Passing durable=True to an LLM operator caches tool calls and LLM outputs mid-task.

18:13 The provider offers three abstraction levels.

21:20 The provider currently requires Airflow 3 — the team is open to adding Airflow 2.11 support if demand is high enough.

24:10 MCP server configs can be stored as Airflow connections.

Resources Mentioned:

Kaxil Naik

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

Pavan Kumar Gopidesu

https://www.linkedin.com/in/pavan-kumar-gopidesu/

Astronomer | LinkedIn

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

Astronomer | Website

https://www.astronomer.io

Experian

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

Apache Airflow

https://www.linkedin.com/company/apache-airflow

Apache Airflow common AI provider docs

https://airflow.apache.org/docs/apache-airflow-providers-common-ai/stable/commits.html

Apache DataFusion

https://datafusion.apache.org/

Pydantic AI

https://pydantic.dev/docs/ai/overview/

Airflow Slack

https://airflow.apache.org/docs/apache-airflow-providers-slack/stable/index.html

Introducing the Common AI Provider: LLM and AI Agent Support for Apache Airflow

https://airflow.apache.org/blog/common-ai-provider/

Thanks for listening to “The Data Flowcast: Mastering Apache Airflow® for Data Engineering and 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.

#Automation #Airflow #MachineLearning

Building AI Debugging Agents Into Airflow DAGs at Jeppesen ForeFlight with Samantha Blaney Cuevas Apr 16, 2026

Show notes

Aviation data pipelines run on strict 28-day publication cycles, and the margin for error is zero. In this episode, we're joined by Samantha Blaney Cuevas, Software Engineer at Jeppesen ForeFlight, to explore how her team orchestrates a complex, time-sensitive data pipeline with Airflow and where AI is starting to fit into that picture.

Key Takeaways:

00:00 Introduction.

04:05 Airflow orchestrates almost all business logic and data transformations across the cycle, with custom timetables built to track busy and slow periods programmatically.

06:10 Cycle-aware sensing tasks handle irregular source deliveries, including duplicates and early or late arrivals, without disrupting the pipeline.

08:07 The two main AI use cases are pipeline debugging and cycle awareness — both designed to reduce the manual overhead of monitoring a complex DAG dependency graph.

09:03 The Data Port agent is a two-task DAG that routes Slack pipeline alerts to either a predefined command list or an AI token, depending on whether the fix is already known.

13:10 AI is still in development at Jeppesen ForeFlight — the team is focused on token efficiency and scoping how much autonomy to give agents across different environments.

15:04 Airflow setup and MCP configuration were straightforward — the harder design work was deciding which environments agents could access across QA staging and production.

17:06 Airflow's skills repo and agent tooling are helping onboard new developers and extend pipeline awareness to analysts who work alongside engineers on the cycle.

19:10 Samantha would like to see single-task retries with different parameters in Airflow — resetting one task without clearing the full pipeline run.

21:05 A future AI use case under consideration is live DAG editing and re-upload within Airflow to make one-off fixes without halting pipeline progress.

Resources Mentioned:

Samantha Blaney Cuevas

https://www.linkedin.com/in/samantha-blaney/

Jeppesen ForeFlight | LinkedIn

https://www.linkedin.com/company/jeppesen-foreflight/

Jeppesen ForeFlight | Website

http://www.foreflight.com

Astronomer Airflow Skills Repo

http://www.github.com/astronomer/airflow-llm-providers-demo

Apache Airflow

https://airflow.apache.org/

#AI #Automation #Airflow

Introducing Airflow 3.2 Apr 09, 2026

Show notes

We introduce Airflow 3.2 and its updates for teams that build and operate data pipelines.

Astronomer’s Head of Customer Education, Marc Lamberti, and Senior Manager of Developer Relations, Kenten Danas, break down what’s new, from asset partitioning to Async Python tasks and DAG versioning. They explore how these updates improve scheduling, performance and observability in production workflows.

Key Takeaways:

00:00 Introduction.

02:10 Airflow 3 architecture separates workers from the metadata database.

03:05 Plugin versioning and UI-based backfills simplify operations.

06:20 Asset partitioning enables granular, partition-level scheduling.

07:15 Triggering DAGs on partitions instead of full datasets.

11:05 Deferrable operators reduce worker slot usage.

12:00 Async operators reduce database pressure and overhead.

14:10 Async improves throughput, not single task speed.

22:20 Inlets and outlets improve asset lineage visibility.

23:00 DAG version markers show changes directly in the UI.

Resources Mentioned:

Marc Lamberti

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

Apache Airflow

https://airflow.apache.org/

Astronomer | LinkedIn

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

Astronomer | Website

https://www.astronomer.io/

3.2 Webinar

https://www.astronomer.io/events/webinars/introducing-airflow-3-2-video

Asset Partitioning Guide

https://www.astronomer.io/docs/learn/airflow-partitioned-runs

Asynchronous Processes Guide

https://www.astronomer.io/docs/learn/deferrable-operators

Release Notes

https://airflow.apache.org/docs/apache-airflow/stable/release_notes.html#airflow-3-2-0-2026-04-07

Provider Registry

https://airflow.apache.org/registry/

#AI #Automation #Airflow #MachineLearning

Reflections on a Decade of Data Engineering at Seattle Data Guy Apr 03, 2026

Show notes

Lessons from the past decade of data engineering reveal how much the ecosystem has changed and what has stayed surprisingly consistent.

In this episode, Benjamin Rogojan, Owner and Data Consultant at Seattle Data Guy, joins us to reflect on how the data engineering landscape has evolved alongside Apache Airflow. We explore when Airflow makes sense as an orchestrator, why batch processing is still dominant and how AI is reshaping the workflows and responsibilities of modern data engineers.

Key Takeaways:

00:00 Introduction.

03:00 Airflow becomes valuable when workflows involve many pipelines, teams and dependencies.

05:00 Data engineers are still focused on making data accessible and aligning work with business needs.

05:30 Batch pipelines remain the most common approach even as real-time use cases grow.

07:45 Many “real-time” requests are actually event-driven batch workflows.

09:00 Airflow replaced many custom-built pipeline systems with built-in dependency management.

11:00 Modern orchestration tools often build on Airflow concepts or differentiate from them.

14:00 AI can assist with writing SQL and pipelines but still requires experienced engineers.

15:30 Organizations are collecting increasingly granular data creating more engineering demand.

19:00 The data stack has shifted rapidly from Hadoop-era systems to modern cloud platforms.

Resources Mentioned:

Benjamin Rogojan

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

Seattle Data Guy

https://www.linkedin.com/company/seattle-data-guy/

Apache Airflow

https://airflow.apache.org

Airflow Summit / Airflow Conference

https://airflowsummit.org

Snowflake

https://www.snowflake.com

HubSpot Data Sharing / APIs

https://developers.hubspot.com

MLflow

https://mlflow.org

#AI #Automation #Airflow

Managing Data Quality and Governance With Airflow at Credit Karma with Ashir Alam Mar 26, 2026

Show notes

Data quality is not optional when you manage credit data at scale.

In this episode, Ashir Alam, Senior Data Engineer at Credit Karma, joins us to share how his team acts as the gatekeeper for credit data ingestion, how they standardize data quality with Airflow and DAG Factory and how they scale safely across thousands of DAGs. We explore how governance, PII protection and orchestration come together inside a modern data platform.

Key Takeaways:

00:00 Introduction.

01:00 Overview of Credit Karma’s products and financial data ecosystem.

02:00 The team acts as gatekeepers for ingesting data from TransUnion and Equifax.

03:00 Why PII handling and controlled downstream access led to adopting Airflow.

04:00 BigQuery as the warehouse and Airflow as the primary orchestrator.

05:00 Why data quality and governance are critical in financial systems.

07:00 Why Airflow was selected: ease of use and unified ETL plus data quality.

09:00 Introduction to DAG Factory and YAML-based DAG generation.

10:00 GitHub executor creates PR-driven DAG workflows with CI checks.

12:00 BigQuery operators, structured checks and custom Slack and PagerDuty alerts.

13:00 Failed checks stop ETL pipelines and trigger notifications.

17:00 Scaling DAG Factory across thousands of DAGs and runtime vs compile-time concerns.

19:00 Future improvements: better defaults, retries and GenAI workflows in Airflow.

Resources Mentioned:

Ashir Alam

https://www.linkedin.com/in/ashir-alam/

Credit Karma

https://www.linkedin.com/company/intuit-credit-karma/

Apache Airflow

https://airflow.apache.org/

DAG Factory

https://github.com/astronomer/dag-factory

BigQuery (Google Cloud)

https://cloud.google.com/bigquery

GitHub

https://github.com/

Slack

https://slack.com/

PagerDuty

https://www.pagerduty.com/

#AI #Automation #Airflow

Open Source Airflow Contributions and Performance Improvements at G-Research with Christos Bisias Mar 19, 2026

Show notes

Modern Airflow isn’t just orchestration. It's a contribution.

In this episode, we explore how open source investment drives real performance gains and deeper observability.

We’re joined by Christos Bisias, Open Source Software Engineer, Apache Airflow at G-Research, to discuss how his team uses Airflow for large-scale data transformations, contributes upstream and improves scheduler throughput and OpenTelemetry support. From trace-level observability to CI-enforced metrics governance and a major scheduler optimization, this conversation spans strategy, engineering and community impact.

Key Takeaways:

00:00 Introduction.

01:20 How G-Research applies machine learning and big data to predict financial market movements.

02:15 Contributing to open source is a business decision.

03:10 Maintaining a fork is costly.

04:30 OpenTelemetry collects metrics, logs and traces to provide deep system visibility.

06:10 Custom spans help identify bottlenecks inside tasks and enable performance optimization.

08:05 OpenTelemetry integration works properly in Airflow 3.0 and above.

10:00 A YAML-based metrics registry with CI enforcement ensures consistency between docs and exported metrics.

12:10 Scheduler throughput improved significantly by applying concurrency limits earlier in the database query.

15:20 Future Task SDK changes may enable language-agnostic DAG authoring beyond Python.

Resources Mentioned:

Christos Bisias

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

G-Research

https://www.linkedin.com/company/g-research/

Apache Airflow

https://airflow.apache.org/

OpenTelemetry

https://opentelemetry.io/

Prometheus

https://prometheus.io/

Grafana

https://grafana.com/

Jaeger

https://www.jaegertracing.io/

#AI #Automation #Airflow

Automating Threat Intelligence Using Airflow with Karan Alang Mar 12, 2026

Show notes

In this episode, Karan Alang, Principal Software Engineer at Versa Networks, joins the conversation to discuss how Airflow can be used to automate threat intelligence in modern cybersecurity environments. He explains the growing scale of cloud computing, the profitability of hacking and the shortage of SOC analysts. Karan also outlines a novel architecture that combines Airflow, XDR, graph databases and LLMs to orchestrate automated threat detection and response.

Key Takeaways:

00:00 Introduction.

05:00 Organizations face massive log volumes and a shortage of SOC analysts.

07:00 The solution integrates Airflow, XDR, Neo4j graph databases and LLMs into one architecture.

08:00 MITRE ATT&CK provides a global framework for mapping tactics and techniques.

11:00 Airflow acts as the orchestration backbone for ingestion graph transformation and LLM workflows.

13:00 Graph databases provide a full relationship view of attackers’ systems and entities.

14:00 LLMs automate mapping activity to MITRE ATT&CK and assign explainable risk scores.

17:00 Traditional signature-based detection allows lateral movement and exfiltration before teams can react.

18:00 End-to-end automation is essential to mitigating modern cybersecurity threats.

20:00 Future opportunities include deeper LLM integration as first-class citizens within Airflow.

Resources Mentioned:

Karan Alang

https://www.linkedin.com/in/karan-alang-4173437

Versa Networks | LinkedIn

https://www.linkedin.com/company/versa-networks

Versa Networks | Website

https://versa-networks.com

Google Cloud Composer (Managed Airflow on GCP)

https://cloud.google.com/composer

Microsoft Defender XDR

https://www.microsoft.com/es-es/security/business/siem-and-xdr/microsoft-defender-xdr

Neo4j (Graph Database)

https://neo4j.com

MITRE ATT&CK Framework

https://attack.mitre.org

#AI #Automation #Airflow #MachineLearning

Using Plugins To Customize Airflow at Ponder Labs with Egor Tarasenko Mar 05, 2026

Show notes

In this episode, we explore how teams scale Apache Airflow in complex environments and what it takes to make orchestration work across many stakeholders. We look at real-world challenges around visibility, ownership and predictability as data platforms grow.

Egor Tarasenko, Data and AI Engineer at Ponder Labs, joins us to share how Ponder Labs customizes Airflow for education organizations using plugins, event-driven architectures and AI-powered tooling. He explains how his team supports large charter school networks and why structure, consistency and extensibility become critical at scale.

Key Takeaways:

00:00 Introduction.

01:21 Ponder Labs helps education organizations bring data from many systems together so it becomes useful for teachers, school leaders and administrators.

03:10 Airflow serves as the backbone for orchestrating ingestion, transformation and reverse ETL across client data platforms.

05:43 Everything is triggered from Airflow to maintain dependency, visibility and a single operational picture.

09:05 Managing hundreds of DAGs requires a focus on structure, visibility and consistency across teams.

09:51 Treating DAGs like APIs helps teams scale without needing deep knowledge of upstream logic.

12:00 Custom plugins like schedule insights help predict DAG run times across layered dependencies.

15:00 AI-powered Airflow chat enables non-technical stakeholders to understand DAG ownership dependencies and cluster activity.

22:06 Migrating plugins to Airflow 3 improves developer experience through cleaner APIs and faster extensibility.

Resources Mentioned:

Egor Tarasenko

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

Apache Airflow

https://airflow.apache.org

dbt

https://www.getdbt.com

Astronomer Astro Platform

https://www.astronomer.io

Egor Tarasenko on Substack

https://egortarasenko.substack.com

#AI #Automation #Airflow

Scaling Airflow at Wix for Analytics and AI with Ethan Shalev Feb 26, 2026

Show notes

Modern data orchestration at scale demands reliability, speed and thoughtful adoption of new tooling. As organizations grow, keeping pipelines efficient while supporting more teams becomes a critical challenge.

In this episode, we’re joined by Ethan Shalev, Data Engineer at Wix, to discuss how Wix operates Airflow at massive scale, migrates to Airflow 3 and uses AI to accelerate development.

Key Takeaways:

00:00 Introduction.

02:13 Wix structures data engineering across multiple product-focused organizations.

03:40 Migrating nearly 8,000 DAGs to Airflow 3 requires careful planning.

04:31 Migration creates an opportunity to remove long-standing legacy Airflow code.

05:32 Internal playbooks and Cursor rules standardize and speed up DAG migrations.

07:39 Airflow 3 introduces backfills, DAG versioning and asset-aware scheduling.

09:16 Deferrable operators reduce scheduler congestion in large Airflow environments.

12:54 AI-generated code still requires review and strong testing practices.

14:52 Moving to managed Airflow reduces operational burden on internal platform teams.

15:57 Improving multi-tenancy and UI personalization remains a key Airflow need.

Resources Mentioned:

Ethan Shalev

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

Wix | LinkedIn

https://www.linkedin.com/company/wix-com/

Wix | Website

https://www.wix.com/

Apache Airflow

https://airflow.apache.org/

Astronomer

https://www.astronomer.io/

Trino

https://trino.io/

Apache Iceberg

https://iceberg.apache.org/

Cursor

https://cursor.sh/

Airflow Summit

https://airflowsummit.org/

#AI #Automation #Airflow

Using Airflow To Orchestrate Billions of Events at Addi with Carlos Daniel Puerto Niño Feb 19, 2026

Show notes

Strong data orchestration is as much about culture and visibility as it is about technology. As data platforms scale, teams need systems that reduce cognitive load while increasing reliability and observability.

In this episode, Carlos Daniel Puerto Niño, Senior Analytics Engineer and Data Analyst at Addi, joins us to share how Addi uses Airflow to support batch orchestration, manage organizational complexity and improve monitoring across its data platform.

Key Takeaways:

00:00 Introduction.

01:25 Changes in company strategy increase data platform complexity over time.

04:00 Centralized data teams help manage organizational and technical change.

06:08 Scalable architectures support growing data volumes and use cases.

09:10 Adopting orchestration tools introduces operational and maintenance challenges.

14:43 Abstraction layers lower technical barriers for onboarding new team members.

15:36 Modularity and visibility improve the reliability of data pipelines.

18:14 Integrated monitoring supports faster incident response and resolution.

22:19 Limited access to orchestration metadata constrains proactive analysis.

Resources Mentioned:

Carlos Daniel Puerto Niño