UrduGPT is an end-to-end LLM engineering project that demonstrates a complete post-training workflow, featuring supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) for the Urdu language. Utilizing the Hugging Face ecosystem, PEFT, and Unsloth, the project delivers highly aligned models for better linguistic performance and enables LLM-as-a-Judge evaluation via Groq. The complete project, including SFT and DPO models, is available for review on GitHub and Hugging Face.