We ❤️ Open Source
A community education resource
Smarter AI workflows: Fine-tuning, declarative ML, and Kubernetes in action
Lessons from the field: How developers can scale AI without losing control.
Shivay Lamba, a developer evangelist at Couchbase, sat down with the All Things Open team to share insights on scaling AI with Kubernetes, fine-tuning large language models (LLMs), and the future of developer productivity.
Read more: Why Kubernetes is essential for AI and open source projects in 2025
Shivay shared how fine-tuning large language models can help organizations get more accurate results without the heavy costs of retraining from scratch. By adjusting model weights with proprietary data, developers can create tools like customer chatbots that respond with context and relevance. This approach makes AI adoption more practical for business-critical use cases.
What is declarative machine learning?
He also introduced the idea of declarative machine learning, where compute resources are assigned step by step across a workflow. Instead of overprovisioning GPUs for every stage, developers can allocate CPUs for lighter tasks like data cleaning, then scale up with GPUs only where needed. This model not only reduces costs but also makes machine learning more approachable for DevOps teams already familiar with Kubernetes.
When asked about scaling AI, Shivay pointed to Kubernetes as the go-to technology. Originally built for software workloads, Kubernetes has become essential for running AI models in distributed environments. With GPU sharding, time slicing, and support for flexible deployment, Kubernetes enables organizations like OpenAI to manage large-scale training clusters. For developers, it means an open source technology that bridges infrastructure and AI with reliability and efficiency.
Key takeaways
- Fine-tuning LLMs offers a cost-effective way to improve accuracy for enterprise use cases.
- Declarative ML helps developers control resource allocation across complex AI workflows.
- Kubernetes has evolved into a powerful platform for scaling AI workloads alongside traditional applications.
In addition to sharing technical insights, Shivay highlighted developer tools like Pieces for Developers (proprietary) and open source projects that help unify workflows across Slack, GitHub, and IDEs. His advice for the community was clear: Keep building and attend All Things Open to stay connected with open source at its best.
More Kubernetes from the podcast
- How curiosity, Kubernetes, and community shaped my open source journey
- Why Kubernetes is essential for AI and open source projects in 2025
More from We Love Open Source
- What is prompt engineering?
- Deep dive into the Model Context Protocol
- How Acorn Labs is rethinking AI adoption with Obot
- Why AI agents are the future of web navigation
- 3 key metrics for reliable LLM performance
The opinions expressed on this website are those of each author, not of the author's employer or All Things Open/We Love Open Source.