10 years after its creation, Airflow is stronger than ever: in last year’s Airflow survey, 81% of users said Airflow is important or very important to their business, 87% said their Airflow usage has grown over time, and 92% said they would recommend Airflow. In this panel discussion, we’ll celebrate a decade of Airflow and delve into how it became the highly recommended industry standard it is today, including history, pivotal moments, and the role of the community. Our panel of seasoned experts will also talk about where Airflow is going next, including future use cases like generative AI and the highly anticipated Airflow 3.0. Don’t miss this insightful exploration into one of the most influential tools in the data landscape.

When developing Machine Learning (ML) models, the biggest challenges are often infrastructural. How do we deploy our model and expose an inference API? How can we retrain? Can we continuously evaluate performance and monitor model drift? In this talk, we will present how we are tackling these problems at the Philadelphia Phillies by developing a suite of tools that enable our software engineering and analytics teams to train, test, evaluate, and deploy ML models - that can be entirely orchestrated in Airflow. This framework abstracts away the infrastructural complexities that productionizing ML Pipelines presents and allows our analysts to focus on developing robust baseball research for baseball operations stakeholders across player evaluation, acquisition, and development. We’ll also look at how we use Airflow, MLflow, MLServer, cloud services, and GitHub Actions to architect a platform that supports our framework for all points of the ML Lifecycle.

In the past 18 months, artificial intelligence has not just entered our workspaces – it has taken over. As we stand at the crossroads of innovation and automation, it’s time for a candid reflection on how AI has reshaped our professional lives, and to talk about where it’s been a game changer, where it’s falling short, and what’s about to shift dramatically in the short term. Since the release of ChatGPT in December 2022, I’ve developed a “first-reflex” to augment and accelerate nearly every task with AI. As a founder and CEO, this spans a wide array of responsibilities from fundraising, internal communications, legal, operations, product marketing, finance, and beyond. In this keynote, I’ll cover diverse use cases across all areas of business, offering a comprehensive view of AI’s impact. Join me as I sort out through this new reality and try and forecast the future of AI in our work. It’s time for a radical checkpoint. Everything’s changing fast. In some areas, AI has been a slam dunk; in others, it’s been frustrating as hell. And once a few key challenges are tackled, we’re on the cusp of a tsunami of transformation. 3 major milestones are right around the corner: top-human-level reasoning, solid memory accumulation and recall, and proper executive skills. How is this going to affect all of us?

Ford Motor Company operates extensively across various nations. The Data Operations (DataOps) team for Advanced Driver Assistance Systems (ADAS) at Ford is tasked with the processing of terabyte-scale daily data from lidar, radar, and video. To manage this, the DataOps team is challenged with orchestrating diverse, compute-intensive pipelines across both on-premises infrastructure and the GCP and deal with sensitive of customer data across both environments The team is also responsible for facilitating the execution of on-demand, compute-intensive algorithms at scale through. To achieve these objectives, the team employs Astronomer/Airflow at the core of its strategic approach. This involves various deployments of Astronomer/Airflow that integrate seamlessly and securely (via Apigee) to initiate batch data processing and ML jobs on the cloud, as well as compute-intensive computer vision tasks on-premises, with essential alerting provided through the ELK stack. This presentation will delve into the architecture and strategic planning surrounding the hybrid batch router, highlighting its pivotal role in promoting rapid innovation and scalability in the development of ADAS features.

NCR Voyix Retail Analytics AI team offers ML products for retailers while embracing Airflow as its MLOps Platform. As the team is small and there have been twice as many data scientists as engineers, we encountered challenges in making Airflow accessible to the scientists: As they come from diverse programming backgrounds, we needed an architecture enabling them to develop production-ready ML workflows without prior knowledge of Airflow. Due to dynamic product demands, we had to implement a mechanism to interchange Airflow operators effortlessly. As workflows serve multiple customers, they should be easily configurable and simultaneously deployable. We came up with the following architecture to deal with the above: Enabling our data scientists to formulate ML workflows as structured Python files. Seamlessly converting the workflows into Airflow DAGs while aggregating their steps to be executed on different Airflow operators. Deploying DAGs via CI/CD’s UI to the DAGs folder for all customers while considering definitions for each in their configuration files. In this session, we will cover Airflow’s evolution in our team and review the concepts of our architecture.

In the realm of data engineering, machine learning pipelines and using cloud and web services there is a huge demand for orchestration technologies. Apache Airflow belongs to the most popular orchestration technologies or even is the most popular one. In this presentation we are going to focus these aspects of Airflow that make it so popular and whether it became the orchestration industry standard.

This talk will explore ASAPP’s use of Apache Airflow to streamline and optimize our machine learning operations (MLOps). Key highlights include: Integrating with our custom Spark solution for achieving speedup, efficiency, and cost gains for generative AI transcription, summarization and intent categorization pipelines Different design patterns of integrating with efficient LLM servers - like TGI/vllm/tensor-RT for Summarization pipelines with/without Spark. An overview of batched LLM inference using Airflow as opposed to real time inference outside of it [Tentative] Possible extension of this scaffolding to Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF) for fine-tuning LLMs, using Airflow as the orchestrator. Additionally, the talk will cover ASAPP’s MLOps journey with Airflow over the past few years, including an overview of our cloud infrastructure, various data backends, and sources. The primary focus will be on the machine learning workflows at ASAPP, rather than the data workflows, providing a detailed look at how Airflow enhances our MLOps processes.

Laurel provides an AI-driven timekeeping solution tailored for accounting and legal firms, automating timesheet creation by capturing digital work activities. This session highlights two notable AI projects: UTBMS Code Prediction: Leveraging small language models, this system builds new embeddings to predict work codes for legal bills with high accuracy. More details are available in our case study: https://www.laurel.ai/resources-post/enhancing-legal-and-accounting-workflows-with-ai-insights-into-work-code-prediction . Bill Creation and Narrative Generation: Utilizing Retrieval-Augmented Generation (RAG), this approach transforms users’ digital activities into fully billable entries. Additionally, we will discuss how we use Airflow for model management in these AI projects: Daily Model Retraining: We retrain our models daily Model (Re)deployment: Our Airflow DAG evaluates model performance, redeploying it if improvements are detected Cost Management: To avoid high costs associated with querying large language models frequently, our DAG utilizes RAG to efficiently summarize daily activities into a billable timesheet at day’s end.

In his presentation, Elad will provide a novel take on Airflow, highlighting its versatility beyond conventional use for scheduled pipelines. He’ll discuss its application as an on-demand tool for initiating and halting jobs, mainly in the Data Science fields, like dataset enrichment and batch prediction via API calls, complete with real-time status tracking and alerts. The talk aims to encourage a fresh approach to Airflow utilization but will also delve into the technical aspects of implementing DAG triggering and cancellation logic. What will the audience learn: Real-life use case of leveraging Airflow capabilities beyond traditional pipeline scheduling, with innovative integration as the infrastructure for ML Platform. Trigger on-demand DAGs through API. Cancel running DAGs. Demonstration of an end-to-end ML pipeline utilizing AWS Sagemaker for batch predictions. Some more Airflow best practices. Join us to learn from Wix’s experience and best practices!

Up until a few years ago, teams at Uber used multiple data workflow systems, with some based on open source projects such as Apache Oozie, Apache Airflow, and Jenkins while others were custom built solutions written in Python and Clojure. Every user who needed to move data around had to learn about and choose from these systems, depending on the specific task they needed to accomplish. Each system required additional maintenance and operational burdens to keep it running, troubleshoot issues, fix bugs, and educate users. After this evaluation, and with the goal in mind of converging on a single workflow system capable of supporting Uber’s scale, we settled on an Airflow-based system. The Airflow-based DSL provided the best trade-off of flexibility, expressiveness, and ease of use while being accessible for our broad range of users, which includes data scientists, developers, machine learning experts, and operations employees. This talk will focus on scaling Airflow to Uber’s scale and providing a no-code seamless user experience

This session reveals an experimental venture integrating OpenAI’s AI technologies with Airflow, aimed at advancing error diagnosis. Through the application of AI, our objective is to deepen the understanding of issues, provide comprehensive insights into task failures, and suggest actionable solutions, thereby augmenting the resolution process. This method seeks to not only enhance diagnostic efficiency but also to equip data engineers with AI-informed recommendations. Participants will be guided through the integration journey, illustrating how AI can refine error analysis and potentially simplify troubleshooting workflows.

Gen AI has taken the computing world by storm. As Enterprises and Startups have started to experiment with LLM applications, it has become clear that providing the right context to these LLM applications is critical. This process known as Retrieval augmented generation (RAG) relies on adding custom data to the large language model, so that the efficacy of the response can be improved. Processing custom data and integrating with Enterprise applications is a strength of Apache Airflow. This talk goes into details about a vision to enhance Apache Airflow to more intuitively support RAG, with additional capabilities and patterns. Specifically, these include the following Support for unstructured data sources such as Text, but also extending to Image, Audio, Video, and Custom sensor data LLM model invocation, including both external model services through APIs and local models using container invocation. Automatic Index Refreshing with a focus on unstructured data lifecycle management to avoid cumbersome and expensive index creation on Vector databases Templates for hallucination reduction via testing and scoping strategies

Every data team out there is being asked from their business stakeholders about Generative AI. Taking LLM centric workloads to production is not a trivial task. At the foundational level, there are a set of challenges around data delivery, data quality, and data ingestion that mirror traditional data engineering problems. Once you’re past those, there’s a set of challenges related to the underlying use case you’re trying to solve. Thankfully, because of how Airflow was already being used at these companies for data engineering and MLOps use cases, it has become the defacto orchestration layer behind many GenAI use cases for startups and Fortune 500s. This talk will be a tour of various methods, best practices, and considerations used in the Airflow community when taking GenAI use cases to production. We’ll focus on 4 primary use cases; RAG, fine tuning, resource management, and batch inference and take a walk through patterns different members in the community have used to productionize this new, exciting technology.

Artificial Intelligence is reshaping the landscape of software development. In this talk, we’ll explore the latest AI breakthroughs improving LLM capabilities for software development use cases. We’ll discuss work and ideas in the field related to Airflow, particularly around model capabilities related to Python, DSLs, and low-resource languages.

During this talk we are going to given an overview of different orchestration approaches (Kubeflow, Ray, Airflow, etc.) when running ML workloads on Kubernetes and specifically we will focus on how to use Kubernetes Batch API and Kubernetes Operators to run complex ML workloads.

In this session Steve Sawyer will discuss a case study for how IBM Data Observability with Databand, collects metadata to build historical baselines, detect anomalies and triage alerts to remediate data quality issues for you data pipelines and warehouses. Additionally, he will provide a Product perspective on the technologies IBM is building to meet the data observability needs across the enterprise, and how it relates to our investments in AI and Data Fabric.

AI workloads are becoming increasingly complex, with unique requirements around data management, compute scalability, and model lifecycle management. In this session, we will explore the real-world challenges users face when operating AI at scale. Through real-world examples, we will uncover common pitfalls in areas like data versioning, reproducibility, model deployment, and monitoring. Our practical guide will highlight strategies for building robust and scalable AI platforms leveraging Airflow as the orchestration layer and AWS for its extensive AI/ML capabilities. We will showcase how users have tackled these challenges, streamlined their AI workflows, and unlocked new levels of productivity and innovation.

Apache Airflow has emerged as the defacto standard for data orchestration. Over the last couple of years, Airflow has also seen increasing adoption for ML and AI use cases. It has been almost four years since the release of Airflow 2 and as a community we have agreed that it’s time for a major foundational release in the form of Airflow 3. This talk will introduce the vision behind Airflow 3, including the emerging technology trends in the industry and how Airflow will evolve in response. Specifically, this will include an overview of the architectural changes in Airflow to support emerging use cases and distributed data infrastructure models. This talk will also introduce the major features and the desired outcomes of the release. Airflow 3 will be a foundational release and therefore this talk will similarly introduce the new concepts being introduced as part of Airflow 3, which may be fully realized in follow-on 3.x releases. The goal of this talk is to raise awareness about Airflow 3 and to get feedback from the Airflow community while the release is still in the development phase.

Nowadays, conversational AI is no longer exclusive to large enterprises. It has become more accessible and affordable, opening up new possibilities and business opportunities. In this session, discover how you can leverage Generative AI as your AI pair programmer to suggest DAG code and recommend entire functions in real-time, directly from your editor. Visualize how to harness the power of ML, trained on billions of lines of code, to transform natural language prompts into coding suggestions. Seamlessly cycle through lines of code, complete function suggestions, and choose to accept, reject, or edit them. Witness firsthand how Generative AI provides recommendations based on the project’s context and style conventions. The objective is to equip you with techniques that allow you to spend less time on boilerplate and repetitive code patterns, and more time on what truly matters: building exceptional orchestration software.

In the last few years Large Language Models (LLMs) have risen to prominence as outstanding tools capable of transforming businesses. However, bringing such solutions and models to the business-as-usual operations is not an easy task. In this session, we delve into the operationalization of generative AI applications using MLOps principles, leading to the introduction of foundation model operations (FMOps) or LLM operations using Apache Airflow. We further zoom into aspects of expected people and process mindsets, new techniques for model selection and evaluation, data privacy, and model deployment. Additionally, know how you can use the prescriptive features of Apache Airflow to aid your operational journey. Whether you are building using out of the box models (open-source or proprietary), creating new foundation models from scratch, or fine-tuning an existing model, with the structured approaches described you can effectively integrate LLMs into your operations, enhancing efficiency and productivity without causing disruptions in the cloud or on-premises.

Ford Motor Company is undergoing a significant transformation, embracing AI and Machine Learning to power its smart mobility strategy, enhance customer experiences, and drive innovation in the automotive industry. Mach1ML, Ford’s multi-million dollar ML platform, plays a crucial role in this journey by empowering data scientists and engineers to efficiently build, deploy, and manage ML models at scale. This presentation will delve into how Mach1ML leverages Apache Airflow as its orchestration layer to tackle the challenges of complex ML workflows that include disparate systems, manual processes, security concerns, and deployment complexities. We will explore the benefits of using Airflow, such as increased efficiency, improved reliability, enhanced scalability, and faster time-to-value. Additionally, we will showcase how Mach1ML utilizes Airflow capabilities to generate reusable templates and streamline environment promotions to further empower Ford’s AI practitioners and accelerate the delivery of cutting-edge AI-powered solutions supporting the next generation of vehicles.

While Airflow is widely known for orchestrating and managing workflows, particularly in the context of data engineering, data science, ML (Machine Learning), and ETL (Extract, Transform, Load) processes, its flexibility and extensibility make it a highly versatile tool suitable for a variety of use cases beyond these domains. In fact, Cloudflare has publicly shared in the past an example on how Airflow was leveraged to build a system that automates datacenter expansions. In this talk, I will share a few more of our use cases beyond traditional data engineering, demonstrating Airflow’s sophisticated capabilities for orchestrating a wide variety of complex workflows, and discussing how Airflow played a crucial role in building some of the highly successful autonomous systems at Cloudflare, from handling automated bare metal server diagnostics and recovery at scale, to Zero Touch Provisioning that is helping us accelerate the roll out of inference-optimized GPUs in 150+ cities in multiple countries globally.

Many organizations struggle to create a well-orchestrated AI infrastructure, using separate and disconnected platforms for data processing, model training, and inference, which slows down development and increases costs. There’s a clear need for a unified system that can handle all aspects of AI development and deployment, regardless of the size of data or models. Join our breakout session to see how our comprehensive solution simplifies the development and deployment of large language models in production. Learn how to streamline your AI operations by implementing an end-to-end ML lifecycle on your custom data, including - automated LLM fine-tuning, LLM evaluation & LLM serving and LoRA deployments

There are 3 certainties in life: death, taxes, and data pipelines failing. Pipelines may fail for a number of reasons: you may run out of memory, your credentials may expire, an upstream data source may not be reliable, etc. But there are patterns we can learn from! Join us as we walk through an analysis we’ve done on a massive dataset of Airflow failure logs. We’ll show how we used natural language processing and dimensionality reduction methods to explore the latent space of Airflow task failures in order to cluster, visualize, and understand failures. We’ll conclude the talk by walking through mitigation methods for common task failure reasons, and walk through how we can use Airflow to build an MLOps platform to turn this one-time analysis into a reliable, recurring activity.