Uploaded December 2025 | Updated November 2026, 14 hours ago
At Ray Summit 2025, Devansh Ghatak from Simplismart shares how KubeRay can unify and simplify the entire lifecycle of GenAI models—from training and fine-tuning to containerization, inference, and large-scale benchmarking—eliminating the messy mix of scripts, ad-hoc orchestration, and manual setup that traditionally slows ML teams down.
He begins by highlighting why managing modern LLM and multimodal pipelines often becomes complicated: distributed jobs, experiment sweeps, fine-tuning runs, packaging steps, and evaluation workflows typically span multiple tools and custom infrastructure. KubeRay addresses this by acting as a single scalable orchestration layer for running distributed model workflows on Kubernetes.
Devansh then walks through real-world examples demonstrating how Ray and KubeRay streamline:
Grid search and hyperparameter exploration
Automated LoRA fine-tuning pipelines
Model containerization and build workflows
Large-scale, parallelized benchmarking runs
He also shows how Simplismart uses Ray to build environment-agnostic, reproducible workflows that run seamlessly across any Kubernetes cluster—bringing consistency, speed, and reliability to the entire GenAI experimentation and deployment process.
Attendees will learn how KubeRay can accelerate distributed ML workflows, reduce operational overhead, and provide a unified platform for managing the complete model lifecycle in modern GenAI stacks.
Liked this video? Check out other Ray Summit breakout session recordings youtube.com/playlist?list=PLzTswPQNepXllnU0C36WtkC0dqkAoDulh
Subscribe to our YouTube channel to stay up-to-date on the future of AI! youtube.com/c/anyscale
🔗 Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
At Ray Summit 2025, Devansh Ghatak from Simplismart shares how KubeRay can unify and simplify the entire lifecycle of GenAI models—from training and fine-tuning to containerization, inference, and large-scale benchmarking—eliminating the messy mix of scripts, ad-hoc orchestration, and manual setup that traditionally slows ML teams down.
He begins by highlighting why managing modern LLM and multimodal pipelines often becomes complicated: distributed jobs, experiment sweeps, fine-tuning runs, packaging steps, and evaluation workflows typically span multiple tools and custom infrastructure. KubeRay addresses this by acting as a single scalable orchestration layer for running distributed model workflows on Kubernetes.
Devansh then walks through real-world examples demonstrating how Ray and KubeRay streamline:
Grid search and hyperparameter exploration
Automated LoRA fine-tuning pipelines
Model containerization and build workflows
Large-scale, parallelized benchmarking runs
He also shows how Simplismart uses Ray to build environment-agnostic, reproducible workflows that run seamlessly across any Kubernetes cluster—bringing consistency, speed, and reliability to the entire GenAI experimentation and deployment process.
Attendees will learn how KubeRay can accelerate distributed ML workflows, reduce operational overhead, and provide a unified platform for managing the complete model lifecycle in modern GenAI stacks.
Liked this video? Check out other Ray Summit breakout session recordings youtube.com/playlist?list=PLzTswPQNepXllnU0C36WtkC0dqkAoDulh
Subscribe to our YouTube channel to stay up-to-date on the future of AI! youtube.com/c/anyscale
🔗 Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute










