# Blog

URL: https://donedeploy.com/blog/

[!AI · Nov 2, 2026 · 4 min readCI/CD for Models and PromptsEvaluation gates, model registries, canary releases and rollback: applying CI/CD discipline to prompts and models in LLM applications.#CI/CD#devops#Kubernetes](https://donedeploy.com/cicd-for-llms/)[!AI · Oct 23, 2026 · 4 min readFine-Tuning or Prompting? A Decision Guide for Small TeamsWhen prompting, RAG or fine-tuning is the right way to adapt a language model, and what each one costs a small engineering team.#Fine-Tuning#LLM#MLOps](https://donedeploy.com/fine-tuning-or-prompting/)[!AI · Oct 15, 2026 · 4 min readAutomating Business Workflows with AI, SafelyWhere AI fits in business process automation, and how human review, guardrails and audit trails keep automated workflows safe.#AI Agents#automation#Business Process](https://donedeploy.com/ai-workflow-automation/)[!AI · Oct 6, 2026 · 4 min readObservability for LLM ApplicationsWhat to measure in an LLM application: tokens, latency, cost per request, OpenTelemetry traces and the quality signals that show when answers get worse.#Grafana#LLM#Monitoring](https://donedeploy.com/llm-observability/)[!AI · Sep 25, 2026 · 4 min readHosted LLM APIs or Self-Hosted Open Models?GPT, Claude and Gemini APIs against self-hosted open models, compared on cost, privacy, latency and the engineering effort each one needs.#cloud#Cost Optimization#LLM](https://donedeploy.com/hosted-or-self-hosted-llm/)[!AI · Sep 16, 2026 · 4 min readRAG in Production: The Boring Parts That MatterIngestion, chunking, refresh, evaluation and access control: the unglamorous work that decides whether a RAG system can be trusted.#automation#Infrastructure#LLM](https://donedeploy.com/rag-in-production/)[!AI · Sep 7, 2026 · 4 min readGPU Nodes on Kubernetes Without Wasting MoneyNode pools, MIG, time-slicing, scale to zero and spot instances: practical ways to keep Kubernetes GPU costs under control for AI workloads.#cloud#Cost Optimization#GPU](https://donedeploy.com/gpu-nodes-kubernetes/)[!AI · Aug 27, 2026 · 4 min readvLLM vs Ollama vs TGI: Choosing an Inference ServerHow vLLM, Ollama and Hugging Face TGI differ in throughput, hardware support and operations, and which one fits your workload.#Infrastructure#Kubernetes#LLM](https://donedeploy.com/vllm-ollama-tgi/)[!AI · Aug 18, 2026 · 4 min readRunning Open Models in Production: A Practical Starting PointWhat it takes to run Llama, Mistral, Qwen, Gemma or DeepSeek in production: picking a model, sizing hardware, serving it and keeping it healthy.#GPU#Infrastructure#Kubernetes](https://donedeploy.com/running-open-models/)[!Devops · Jun 11, 2026 · 3 min readSimpler Infrastructure, Faster Engineering TeamsSimpler infrastructure helps engineering teams move faster, deploy more efficiently, and focus on building products.#automation#CI/CD#cloud](https://donedeploy.com/simpler-infrastructure-faster-engi/)[!Devops · May 28, 2026 · 2 min readThe Hidden Cost of Manual InfrastructureManual infrastructure creates hidden operational costs through inconsistent deployments, repeated processes, and growing engineering complexity.#automation#cloud#Deployments](https://donedeploy.com/the-hidden-cost-of-manual-infrastructure/)[!Devops · May 14, 2026 · 4 min readInternal Developer Platforms (IDP): Build or Buy?Should you build your own Internal Developer Platform (IDP) or buy one? Compare costs, setup time, and focus on your core product with DoneDeploy.#automation#CTO Guide#Development](https://donedeploy.com/internal-developer-platforms-idp-build-or-buy/)