Uploaded October 2024 | Updated September 2026, 2 weeks ago

---

Watch Steve Han, Wei Zeng, and Yiqing Wang from Roblox present their experiences in leveraging Ray for efficient batch inference in machine learning at Ray Summit 2024. The talk focuses on Roblox's approach to scaling their ML infrastructure, particularly in light of recent advancements in multimodal language models.

The speakers discuss the integration of multimodal models into vLLM, an open-source project that has garnered significant interest from the community. They explore the technical challenges encountered during this process and share key insights gained. The presentation offers a practical perspective on implementing cutting-edge ML technologies in a large-scale gaming platform, providing attendees with valuable lessons applicable to their own ML scaling efforts.

--

Interested in more?
- Watch the full Day 1 Keynote: youtu.be/jwZHJthQvXo
- Watch the full Day 2 Keynote youtu.be/Lury2ad6KG8

--

🔗 Connect with us:
- Subscribe to our YouTube channel: youtube.com/@anyscale
- Twitter: https://x.com/anyscalecompute
- LinkedIn: linkedin.com/company/joinanyscale
- Website: anyscale.com