Bangalore Apache Airflow® Meetup at Uber

Meet up with fellow Airflow users at Uber's Bangalore Office for an afternoon of awesome talks and networking.

This is the perfect opportunity to get to know like-minded Airflow users, partners, and company experts, and learn how customers are utilizing Airflow within their organizations.

🔥 Agenda

• 2:30–2:45 pm – Registration & Welcome
• 2:45–3:10 pm - Talk 1: Standardizing Airflow Pipeline Development Experience
• Speakers: Dinesh Jagannathan, Tech Lead Manager at Uber / Sambit Acharya, Senior Software Engineer at Uber
• Currently, Airflow pipelines are authored as raw Python code, lack environment isolation, rely on ad-hoc testing, and deploy without CI/CD gates. At Uber, we standardized a single, declarative, environment-aware authoring and deployment model with automated pre-prod gating. Without altering core pipeline functionality, the new experience shifts quality checks left, enables safe code promotion, and significantly boosts developer productivity.
• 3:15–3:40 pm - Talk 2: Microservice-Grade Delivery and Release for Airflow DAGs at Uber
• Speakers: Piyush Maheshwari, Sr. Software Engineer at Uber / Sameer Raj, Software Engineer at Uber
• At Uber, we're rethinking DAG delivery with dependency-aware slim bundles that replace 2 GB monorepo artifacts, and safer DAG releases with per-DAG version pinning and deployment gating. Together, these enable regression gates, controlled promotion, and automated rollback at scale.
• 3:40–4:00 pm – Coffee break
• 4:00–4:25 pm – Talk 3: 𝗪𝗵𝗲𝗿𝗲 𝗮𝗿𝗲 𝘆𝗼𝘂 𝘀𝗮𝘃𝗶𝗻𝗴 𝘀𝘁𝗮𝘁𝗲 𝗶𝗻 𝗔𝗽𝗮𝗰𝗵𝗲 𝗔𝗶𝗿𝗳𝗹𝗼𝘄?
• Speaker: Amogh Desai, Sr. Software Engineer at Astronomer
• Airflow was designed around stateless tasks: same inputs, same result, no memory of the last attempt. That is what makes retries and backfills safe to run. However, real-world pipelines often require memory. Where did I get to? Which job did I already launch? Which files are processed? That state is part of the work, and it has to live somewhere. This talk covers the common patterns where state is genuinely needed and explores where that state should actually live in Airflow. Chances are, you have already solved this in your own way. Come and share how you approached it!
• 4:30–4:55 pm – Talk 4: When Airflow Tasks Have Physical Consequences
• Speaker: Shubham Raj, Sr. Engineer, Qualcomm
• Most Airflow tasks move data, some move robots. Airflow is best known for orchestrating data pipelines, but what happens when workflows need to interact with real-world hardware? In this talk, I will share how we use Apache Airflow to automate semiconductor validation labs, coordinating robotic handlers, test systems, and silicon devices across multiple locations.
• We'll also look at how Airflow's Edge Execution model allows a central deployment to securely run tasks close to the hardware, making it possible to manage distributed labs from a single control plane.
• 5:00–5:25 pm – Talk 5: Streamlining Your Airflow Upgrade: Essential Tools for Migrating from 2.x to 3
• Speaker: Ankit Chaurasia, SWE, Apple
• Airflow 3 has been out for over a year, but many teams are still on 2.x because nobody wants to hunt down every breaking change by hand. This talk shows how to automate most of that work. We'll use airflow config update to find and fix removed, renamed, and deprecated settings in your airflow.cfg. Then we'll use Ruff's AIR rules to scan your DAG code for removed APIs, imports that moved to providers, and missing schedule arguments. We will talk about how we can leverage AI models to enable the upgrade. Live demos included.
• 5:30–6:30 pm – Snacks and networking

Please note that confirmed attendees will receive a confirmation message from us and Uber's office location details for the meetup.