![[Uber Seattle] Michelangelo (MA) Learners and Transformers @UberEngineering](https://i.ytimg.com/vi/I07n5qHnrpo/hqdefault.jpg)
Uploaded May 2019 | Updated September 2026, 2 weeks ago
During this April 2019 meetup, Uber engineer Mingshi Wang discusses the architectures of Michelangelo (MA) learners and transformers. A Michelangelo (MA) Learner is a workflow authoring framework for Uber's Michelangelo machine learning platform that allows Uber data scientists, researchers, and engineers to solve complex machine learning problems with customized workflows. An MA Learner provides Python SDK's that make it easy to write MA models on Jupiter notebooks while hiding the underlying complexities of distributing the machine learning jobs to different computing environments. At Uber, machine learning models are represented by pipelines composed of MA transformers. The data preparation and training-- processes that involve one or more estimators-- produce a trained pipeline with MA transformers. The trained pipelines are persisted for subsequent usage by batch and online predictions.
During this April 2019 meetup, Uber engineer Mingshi Wang discusses the architectures of Michelangelo (MA) learners and transformers. A Michelangelo (MA) Learner is a workflow authoring framework for Uber's Michelangelo machine learning platform that allows Uber data scientists, researchers, and engineers to solve complex machine learning problems with customized workflows. An MA Learner provides Python SDK's that make it easy to write MA models on Jupiter notebooks while hiding the underlying complexities of distributing the machine learning jobs to different computing environments. At Uber, machine learning models are represented by pipelines composed of MA transformers. The data preparation and training-- processes that involve one or more estimators-- produce a trained pipeline with MA transformers. The trained pipelines are persisted for subsequent usage by batch and online predictions.
![[Uber Seattle] Michelangelo (MA) Learners and Transformers](https://i.ytimg.com/vi/I07n5qHnrpo/mqdefault.jpg "[Uber Seattle] Michelangelo (MA) Learners and Transformers
During this April 2019 meetup, Uber engineer Mingshi Wang discusses the architectures of Michelangelo (MA) learners and transformers. A Michelangelo (MA) Learner is a workflow authoring framework for Ubers Michelangelo machine learning platform that allows Uber data scientists, researchers, and engineers to solve complex machine learning problems with customized workflows. An MA Learner provides Python SDKs that make it easy to write MA models on Jupiter notebooks while hiding the underlying complexities of distributing the machine learning jobs to different computing environments. At Uber, machine learning models are represented by pipelines composed of MA transformers. The data preparation and training processes that involve one or more estimators produce a trained pipeline with MA transformers. The trained pipelines are persisted for subsequent usage by batch and online predictions.")
![[Uber Marketplace] Marketplace Experimentation Vivek Trehan](https://i.ytimg.com/vi/IR000RqN7pw/mqdefault.jpg "[Uber Marketplace] Marketplace Experimentation Vivek Trehan
During a May 2018 meetup, engineer manager Vivek Trehan discusses how we can improve features and services on our platform using our experimentation platform.
Learn more about our experimentation platform: https://eng.uber.com/xp/")
![[MoneyCon 2019] Books: Scalable, Flexible, and Immutable Storage of Squares Financials](https://i.ytimg.com/vi/I_Pt_i3ntGw/mqdefault.jpg "[MoneyCon 2019] Books: Scalable, Flexible, and Immutable Storage of Squares Financials
The domain of Payments, Finance, and more generally FinTech, is a fast-growing industry that reached record global investment of $111.8B in 2018. The underlying technology that powers this incredible growth is also evolving rapidly. Hosted by Uber, MoneyCon 2019 brought together engineers from leading tech companies to present on the latest topics in payments engineering.
In this talk during MoneyCon 2019, Squares Anthony Bishopric discusses how Square applied a novel cryptographic concept and classic double entry accounting principles to process and disburse hundreds of millions of dollars per day.")
![[Destination:Web] Script Alert 1](https://i.ytimg.com/vi/IwuNdhmk5DA/mqdefault.jpg "[Destination:Web] Script Alert 1
During this July 2019 meetup in San Francisco, Alex Sexton from Stripe discusses how the web ecosystem of 2019 leaves plenty of vulnerable surface area to worry about.")
![[Distributed Tracing NYC] Distributed Transactions Parallel Commits](https://i.ytimg.com/vi/Ize6coEqF98/mqdefault.jpg "[Distributed Tracing NYC] Distributed Transactions Parallel Commits
During a September 2019 meetup hosted at Ubers NYC Engineering Office, Peter Mattis, co-founder of Cockroach Labs, discusses how distributed transactions work and explains the benefits of parallel commits through foundational computer science theory.")
![[Visualization Nights] SharedStreets.io](https://i.ytimg.com/vi/J5FxlVnI2DE/mqdefault.jpg "[Visualization Nights] SharedStreets.io
During Ubers September 2019 Visualization Nights meetup, SharedStreet engineer Morgan Herlocker introduces SharedStreets and discusses how their open source software enables public-private collaboration around and the exchange of transportation data. In this talk, he also demonstrates how to link Uber Movement data with OpenStreetMap and load the data set for visualization in Kepler.gl")
![[Uber Open Source] Petastorm: Uber ATGs Data Access Library for Deep Learning Yevgeni Litvin](https://i.ytimg.com/vi/J5SILwWOktM/mqdefault.jpg "[Uber Open Source] Petastorm: Uber ATGs Data Access Library for Deep Learning Yevgeni Litvin
During a March 2019 meetup hosted at Ubers SF office, Uber ATG staff software engineer Yevgeni Litvin discusses how data produced and managed by Big Data systems like Apache Spark and Apache Hive cannot be directly consumed by Deep Learning systems like TensorFlow and PyTorch.")
![[Destination:Web] JavaScript Errors Demystified](https://i.ytimg.com/vi/KpeftlLF99c/mqdefault.jpg "[Destination:Web] JavaScript Errors Demystified
During a May 2019 meetup hosted at Uber’s San Francisco office, Sentry engineer, Ben Vinegar discusses Collecting JavaScript errors from the browser: its harder than you think. Join him down the rabbit hole of normalizing stack traces, managing CORS, and unwrangling uncaught promise errors.")
![[Uber Mobility] Memory Leak Hunt: LeakCanary - Pierre-Yves Ricau](https://i.ytimg.com/vi/KwArTJHLq5g/mqdefault.jpg "[Uber Mobility] Memory Leak Hunt: LeakCanary - Pierre-Yves Ricau
During our May 30, 2018 Uber Mobility meetup, Square engineer Pierre-Yves Ricau walks through the mental process involved in understanding a leak trace with a live app demo.")
![[Uber Open Summit 2018] Large-Scale Geospatial Visualization in the Browser with deck.gl](https://i.ytimg.com/vi/L2ShgAe85Fg/mqdefault.jpg "[Uber Open Summit 2018] Large-Scale Geospatial Visualization in the Browser with deck.gl
Visualization and interaction with large-scale geospatial datasets is crucial to understanding the pulsing lives and transportation in our cities. deck.gl is a WebGL-powered data visualization framework and the foundation to many real-time, high-performance web applications that power Uber’s business and beyond. During this hands-on workshop, engineer Xiaoji Chen teaches Uber Open Summit 2018 attendees about deck.gl’s core concepts and capabilities via real-world examples.")
![[Destination:Web] Accessibility for Modern Web Apps](https://i.ytimg.com/vi/LCtaccHLjUA/mqdefault.jpg "[Destination:Web] Accessibility for Modern Web Apps
Asynchronous iterators combine the async and await keywords with generator functions, providing an elegant way to express streaming operations in modern JavaScript applications. In this Destination:Web talk, Stripes Paul Ryan explores a real-world use case for async iterators and demonstrates how to properly use the feature with cancellation.")
