AAIF x Uber: Open Source at Uber and Polars

The platform behind every Uber ETA and one of the fastest DataFrame libraries around, both open source. Come see how they work. Since 2016, Michelangelo AI has been the platform Uber’s engineers use to build, train, deploy and monitor machine learning, from ETAs and pricing to fraud detection. It grew from tree models to deep learning to LLMs, and today it also carries Uber’s agent workloads. Uber has now released its core as Open Source under the Apache 2.0 license. Polars has been built in the open since 2020. With a Rust query engine and a Python API on Apache Arrow, it has become a go-to DataFrame library for data teams, backed by an active contributor community and the Polars company. For this edition, AAIF Community Amsterdam, Uber and Polars are spending an evening on open source. We’ll cover how both projects are built, what the teams have learned from developing them in the open, and what they can mean for your own platform. Expect architecture, trade-offs and real production numbers, not a product pitch. What to expect How Michelangelo AI is built: Kubernetes-native, Python-first pipelines with Uniflow, and federated compute across clouds What it takes to serve ML at Uber’s scale How Uber runs LLMs and agents on top of its ML platform What makes Polars fast: lazy evaluation, query optimisation and its Rust engine The Polars open-source journey: growing a community and a company around an open project How to get started with and contribute to both projects Drinks, food and good company Who should come ML engineers, platform engineers, data engineers, data scientists and technical leads who build or run ML, data or agent infrastructure, and anyone curious how open-source infrastructure is designed, built and shared. All experience levels are welcome. Schedule 17h30: Doors 18h00: Opening Remarks 18h05: Lightning talk: Agentic Open Source, Chris Howard (Uber) 18h15: Introducing Michelangelo AI & Demo, Melda Salhab and Qiushen Wang (Uber) 18h45: Chiel Peters (C