Ray Summit 2024 Keynote Day 2 | Where Builders Create the AI Future

Watch the Ray Summit 2024 Day 2 keynote for a compelling look into the transformative power of AI across various industries, featuring insights from tech giants and innovative startups alike. This video showcases how Ray, as the AI Compute Engine, and Anyscale are enabling organizations to break through the AI Complexity Wall and drive groundbreaking advancements in AI applications, from fleet management to drug discovery and e-commerce.
--
The keynote opens with Robert Nishihara and Keerti Melkote recapping Day 1's announcements • Ray Summit 2024 Keynot... , highlighting how Anyscale improves and enhanced Ray performance, with:
- Ray Data enhancements making it 4.5x faster than open source versions
- Ray Serve optimizations offering 56% speed improvements
- Ray Train advancements reducing costs by up to 60%
- 5.1x faster node autoscaling
- Up to 6x lower costs for LLM batch inference compared to AWS Bedrock
They reinforce how these improvements address the growing AI Complexity Wall, a challenge arising from unprecedented scale, multimodal data processing, sophisticated post-training techniques, AI-powered optimization, and the rise of compound AI and agentic systems. Ray, as the AI Compute Engine, is presented as the solution that supports any AI/ML workload, handles diverse data types and model architectures, optimizes resource allocation, and scales seamlessly while abstracting complexity from developers.
--
Following their presentation, we invited a host of external speakers and industry experts to the stage to share their AI journeys.
9:00 John Bicket, CTO of Samsara, details how they're leveraging Ray and Anyscale to build agentic systems supporting their fleet workers, emphasizing Ray's versatility in orchestration, model training, and inference across backend and edge deployments.
29:17 Stephen MacKinnon from Recursion shares their journey in revolutionizing drug discovery with Anyscale, processing 1.3PB of data on 4500 GPUs in just 5 days-a task that previously took months. He highlights how Anyscale freed their data scientists to focus on data alignment and transformation rather than infrastructure concerns.
51:27 Instacart's Co-Founder Brandon Leonardo discusses Instacart's AI evolution, from customer service chatbots to advanced multimodal AI applications. He notes that Anyscale's infrastructure, powering batch vLLM for catalog attribute extraction, delivers faster results at one-tenth the cost of alternatives.
1:08:27 The keynote concludes with a fireside chat between Robert Nishihara and Kevin Weil, CPO of OpenAI. They explore the landscape of AI models, from ChatGPT to OpenAI's latest O1 model, and discuss the impact of open-source LLMs on the industry.
Watch this keynote to get a panoramic view of AI's current capabilities and future potential, underscoring Ray and Anyscale's pivotal role in driving innovation across diverse sectors.
--
Liked this video? Check out the Ray Summit breakout session recordings!
--
🔗 Connect with us:
Subscribe to our RUclips channel: / @anyscale
Twitter: x.com/anyscale...
LinkedIn: / joinanyscale
Website: www.anyscale.com