Koba: Disaggregated Data Loading for Multimodal Training at Luma AI | Luma AI | Ray Summit 2026 @anyscale
Koba: Disaggregated Data Loading for Multimodal Training at Luma AI | Luma AI | Ray Summit 2026  @anyscale
Uploaded September 2026 | Updated November 2026, 2 days ago
As models and data grow, the multimodal training bottleneck increasingly comes from the surrounding input work rather than the training step itself: decoding, transformation, and tokenization contending with gradient computation for accelerator time.

At Ray Summit 2026, Minghao Xie and Haoran Cai from Luma AI present Koba, their multimodal data-loading framework, and Kojo, a Ray-based substrate that disaggregates stateless preprocessing stages onto managed actor pools while preserving the existing training loop. The result: compute that scales on its own pool independently of the trainer, fault isolation so a preprocessing failure no longer cascades into training, and reclaimed accelerator capacity returned to the gradient path.

You'll leave with a reusable pattern for relocating stateless, out-of-step computation off the training GPUs and onto Ray, plus the operational lessons of running it at scale.

Liked this video? Check out other Ray Summit breakout session recordings: youtube.com/playlist?list=PLZNaJyYZRPdo

Subscribe to our YouTube channel to stay up-to-date on the future of AI: youtube.com/c/anyscale

đź”— Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
Website: anyscale.com
Koba: Disaggregated Data Loading for Multimodal Training at Luma AI | Luma AI | Ray Summit 2026A Compiler for Christmas | Brendan Burns #raysummitTowards Clinically Grounded LLMs | SonderMind | Ray Summit 2026Scaling LLM Post-Training at Character.AI | Ray Summit 2025Airbnbs LLM Evolution: Fine-Tuning with Ray | Ray Summit 2024Webinar: Getting Started with Distributed Training at ScaleHighly Compressed Gemma Models with Custom Quantization-Aware Training | DeepMind | Ray Summit 2026Multi-Modal Data Processing Pipelines for AI Systems | Anyscale | Ray Summit 2026Ray Across the Stack Fine Tuning Qwen and Serving 3D Generation at Autodesk with AWS Trainium and InBuilding AV 3.0 for Autonomous Trucking | Felix Heide (Torc Robotics) | Ray Summit 2026Scaling PyTorch Training with Ray Data on Ubers Michelangelo AI Platform | Uber | Ray Summit 2026How Alibaba Cloud Accelerates AI Pipelines with AnalyticDB Ray | Ray Summit 2025
Anyscale |

Koba: Disaggregated Data Loading for Multimodal Training at Luma AI | Luma AI | Ray Summit 2026

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER