New Ray and Anyscale platform features and capabilities unveiled to 12,000 Ray Summit registrants
Anyscale, the company behind the open source project Ray, kicked off its second annual Ray Summit today. The three-day virtual event features an impressive speaker lineup of leading machine learning (ML), artificial intelligence (AI) and distributed system experts from Uber, Robinhood, Google, Amazon, Shopify, Visa, Dendra Systems, Wildlife Studios, and more.
In its second year, Ray Summit will welcome nearly 12,000 Ray users and enthusiasts — developers, ML engineers, and researchers — from over 140 countries, doubling the signups from the 2020 event. More than 50 talks will showcase real-world examples of large-scale, distributed ML and data applications powered by Ray, as well as the latest innovations in the Ray ecosystem.
“We are thrilled by the momentum in the Ray community, not only in user growth and adoption, but also in the ecosystem and the diversity of use cases being powered by Ray,” said Robert Nishihara, co-founder and CEO, Anyscale, and one of the co-creators of Ray. “Many of the most popular ML libraries such as XGBoost, Horovod, Dask, Airflow, and more now integrate with Ray, enabling it to become the unified distributed framework for end-to-end ML workflows. It is very exciting to see this evolution reflected in many of the talks at this year’s Ray Summit.“
Session highlights include:
- McKinsey / QuantumBlack describes how they used AI and Ray to design a world-class sailor to speed up hydrofoil designs, and partnered with Emirates Team New Zealand to win the 36th America’s cup.
- Uber, Shopify, Robinhood discuss how they are building centralized ML platforms, and why they are using Ray as the underlying compute substrate.
- Ant Group describes how using Ray as the distributed computing foundation for their Fusion Engine is helping them efficiently scale a variety of business applications from risk management to growth marketing.
- Dendra Systems presents how they are using ML and large-scale aerial imagery to accelerate ecosystem restoration and rehabilitation at scale.
- Wildlife Studios shares how they improved their offer recommendation engine using reinforcement learning, and increased player lifetime value.
“The appeal of Ray is evident in its broad adoption. We see large companies like Uber building sophisticated machine learning platforms on top of Ray, as well as smaller companies with limited engineering resources, building scalable end-to-end machine learning applications,” said Ben Horowitz, general partner and co-founder at Andreessen Horowitz. “Anybody who wants to build a scalable machine learning application will want to consider using Ray.”
Anyscale will unveil several new capabilities on the Anyscale platform that will enable developers to go from prototyping on their laptop to experimenting at scale in the cloud to deploying in production, all with minimal code changes. These new capabilities further Anyscale’s goal of being the best platform to create, run, and manage Ray applications. Feature highlights include:
- New client and dependency syncing enables developers to scale from their laptop to the cloud with no code changes
- App packaging automatically bundles all code and library dependencies, enabling seamless migration to production
- SDK and REST APIs give programmatic access to clusters, enabling integration with existing automation workflows like CI/CD pipelines
“Ray has enabled us to develop, test, and deploy a new offer recommendation engine based on reinforcement learning in under five months, and it’s already shown increased player lifetime value,” said Emiliano Castro, principal data scientist at Wildlife Studios. “We are data scientists, not distributed systems experts. Without Ray, it would have taken us significantly longer to bring this new idea to market. The scale, functionality, and support that Anyscale provides accelerates our time to market, enabling us to launch new experiences sooner.”
For more such updates and perspectives around Digital Innovation, IoT, Data Infrastructure, AI & Cybsercurity, go to AI-Techpark.com.