samuelstolicny/claudie-expert-gemma4-31b

VISIONPricing:Input $0.48 / Cached $0.1 / Output $1.44Concurrent Unit Cost:2Model Size:31BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The samuelstolicny/claudie-expert-gemma4-31b is a 31 billion parameter Gemma 4-based instruction-tuned language model, fine-tuned by samuelstolicny. It specializes in providing expert knowledge on Claudie, an open-source platform for managing multi-cloud and hybrid-cloud Kubernetes infrastructure. This model excels at answering technical questions related to Claudie's architecture, microservices, InputManifest CRD authoring, and debugging common issues. With a context length of 32768 tokens, it is designed for in-depth technical support and knowledge retrieval within the Claudie ecosystem.

Loading preview...

Claudie Expert - Gemma 4 31B

This model is a fine-tuned version of the Gemma 4 31B-IT base model, specifically specialized in the Claudie platform. Claudie is an open-source platform developed by Berops for declaratively managing multi-cloud and hybrid-cloud Kubernetes infrastructure across various providers like AWS, Azure, GCP, and OpenStack.

Key Capabilities

  • Claudie-Specific Knowledge: Fine-tuned on approximately 8,000 Claudie-specific Q&A conversations.
  • Comprehensive Coverage: Understands Claudie's 8 microservices (Manager, Terraformer, Ansibler, Kube-Eleven, Kuber, Claudie-Operator, Autoscaler-Adapter).
  • Infrastructure Management: Proficient in InputManifest CRD authoring for multi-cloud, GPU, and autoscaling Kubernetes clusters.
  • Debugging Expertise: Capable of assisting with debugging common Claudie issues, including stuck states, NATS consumer lag, Terraform state locks, and WireGuard problems.
  • Architectural Understanding: Knows about gRPC service communication, state machines, reconciliation loops, and data flow within Claudie's architecture.

Training Details

The model was trained using LoRA bf16 (rank 64, alpha 128, all-linear) on a dataset of 8,012 synthetic Q&A conversations derived from the Claudie codebase and documentation. The base model used was unsloth/gemma-4-31B-it.

Good For

  • Developers and operators working with Claudie who need quick technical answers.
  • Automated support systems for Claudie users.
  • Generating InputManifest configurations or troubleshooting Claudie deployments.

Finetunes
2 models