Skip to main content
AmigoFish

Pods in the Wild: Reinforcement Learning... Sunil Arora, Google & Ravin Kumar, Google Deepmind (ASL)

From CNCF [Cloud Native Computing Foundation] · · 9 min

The player loads from YouTube only when you click. It uses YouTube's privacy-enhanced mode.

If the video does not play, the owner may have turned off embedding. Watch it on YouTube instead.

Watch on YouTube

Not rated yet

Description

Connect with the cloud native community and explore CNCF events: https://kubecon.io

Pods in the Wild: Reinforcement Learning on Kubernetes with OpenRL - Sunil Arora, Google & Ravin Kumar, Google Deepmind (ASL)

What do wild dolphins, open-weight Gemma models, and Kubernetes have in common?

Pods, of course!

By training DolphinGemma on decades of marine acoustic data, researchers are actively decoding dolphin communication. This reflects a broader industry shift: transformative AI pairs domain-specific open weights with specialized data. Far from replacing frontier models, winning architectures unite massive foundation LLMs for broad reasoning with lightweight, task-tuned models for targeted execution. While frontier models set the bar for general intelligence, the next major operational leap belongs to everyday builders: empowering practitioners to fine-tune open-weight models that outperform massive frontier baselines on domain-specific tasks at a fraction of the cost, footprint, and latency.

Historically, reinforcement learning (RL) was trapped inside elite research labs and brittle, complex runtimes. A new CNCF sandbox project, OpenRL fundamentally changes this paradigm by democratizing RL directly on Kubernetes. Now, a builder can articulate a task, point to a dataset, and use CNCF technologies to guide an open-weight model toward task mastery without requiring deep machine learning expertise.

Back to CNCF [Cloud Native Computing Foundation] · Report a problem