Anyscale

Amazing to see Nvidia building with Ray.

Nvidia just released a large-scale video curation pipeline at GTC built with Ray. Why not use existing batch processing systems? A couple technical reasons. 1. Supporting heterogeneous resources: batch processing systems that process one stage at a time will typically underutilize GPUs when the CPU stage is running (see figure). 2. Latency: reading & writing intermediate data products between stages slows things down. They process ~1M hours of 720p video data on 2k H100 GPUs in one day. https://lnkd.in/eCe37rsG

I first became aware of "generative AI" with the GAN paper in 2014 (written by Ian Goodfellow and others). Back then, generative AI could generate a low-res number or something that looked kind of like a face. For anything more complex, you were lucky to get something that looked like an impressionist painting (cropped to the size of a postage stamp). In light of that, it's amazing how far Runway has pushed the state of the art.

🚨 Enhance Your Ray Skills with Anyscale Training 🚨 We have two upcoming hands-on courses to help you effectively build and scale AI and ML workloads with Ray: ⭐ Intro to Ray and Anyscale (April 7-8) – A two-day (half-day sessions) course for those new to Ray. Learn model training, hyperparameter optimization, data processing, and scalable deployment. ⭐ Ray for Practitioners (April 7-11) – A five-day (half-day sessions) deep dive for experienced developers. Cover advanced topics like distributed data processing, large-scale ML execution, and deployment strategies. Both courses are guaranteed to run! ✅ 🔥 Save 50% with promo code rayai50 (limited-time offer) 📅 Sign up today: 👉 Intro to Ray & Anyscale: https://lnkd.in/gsW4xZQb 👉 Ray for Practitioners: https://lnkd.in/dXuJ9SFf

Runway is pushing the limits of generative AI – proving that innovation accelerates when infrastructure gets out of the way. With Anyscale, they scale effortlessly, freeing their team to focus on building cutting-edge AI for media creation. "Anyscale enables us to push the boundaries of what’s possible in generative AI by giving us the flexibility to scale workloads seamlessly. This removes the risk around our infrastructure and allows our team to focus on innovation rather than infrastructure bottlenecks." – Anastasis Germanidis, Co-Founder & CTO @ Runway Learn how Runway scales AI-driven media creation, powered by Anyscale: https://lnkd.in/g9DXthNk

Join us on March 27th for our next Ray meet-up, co-hosted with Daft! We’ll cover scaling data processing & ML training with Daft + Ray! Hear from Daft Software Engineer Desmond Cheong & Anyscale Product Manager Ricardo Decal on building AI applications from local to production. Network with fellow AI enthusiasts and grab some pizza! 📍 Sign up today: https://lu.ma/u5p41kte

🔹 Level Up Your Ray Skills with Anyscale Public Courses! 🔹 We’ve got two upcoming hands-on training courses designed to help you build and scale AI and ML workloads with Ray: ⭐ Introduction to Ray and Anyscale (April 7-8) 2 half-day course. Ideal for those looking to get started with Ray, covering model training, HPO, data processing, and scalable deployment. ⭐ Ray for Practitioners (April 7-11) – 5 half-day course. A deep dive for experienced developers to master Ray’s advanced capabilities, including distributed data processing, scalable ML execution, and real-world deployment strategies. ✅ Both courses are guaranteed to run! 🔥 Use promo code rayai50 for 50% off – limited-time offer! 📅 Don’t miss out—secure your spot today: 👉 Intro to Ray & Anyscale: https://lnkd.in/gsW4xZQb 👉 Ray for Practitioners: https://lnkd.in/dXuJ9SFf

⭐ Ray Public Courses Alert ⭐ Newly launched 🚀 Ray for Practitioners course runs April 7-11! This in-depth, 5-day training is designed for experienced developers looking to deepen their expertise in Ray’s advanced capabilities. Topics include Ray core concepts, distributed data processing, scalable execution of end-to-end ML workloads, and real-world deployment strategies. ⚡ Guaranteed to run! ⚡ 🔥 Use promo code rayai50 for 50% off—limited time only. Sign up now https://lnkd.in/dXuJ9SFf

⚡Announcing Anyscale's Technical Webinar Series⚡ The series is a monthly program designed to provide engineers, data scientists, and ML practitioners with practical, hands-on experience in using Ray for distributed AI workloads. Each session is structured to address real-world AI use cases and distributed computing challenges, aiming to equip participants with the skills needed to optimize large-scale workloads effectively. 🗓️Full Webinar Series Schedule: https://lnkd.in/g47s4q6j 🔹Next Webinar in Our Series – March 26: Introduction to Ray Data: Processing Large Datasets on Heterogeneous Clusters 🔹 Learn how Ray Data enables distributed data processing and how to optimize performance across CPUs and GPUs. This hands-on session will cover Ray Data's architecture, core concepts, and practical applications in production ML pipelines, using real-world datasets. Register now 👉 https://lnkd.in/gYWueasP

Join us on March 27th for our next Ray meet-up, co-hosted with Daft, where we’ll explore scaling data processing and ML training with Daft + Ray! Hear from Daft Software Engineer Desmond Cheong and @anyscalecompute Product Manager Ricardo Decal as they share insights on building AI applications from local production 🚀 Network with fellow AI enthusiasts and enjoy some pizza! 🎉 🍕 Don’t miss out—sign up now! https://lu.ma/u5p41kte