The LLM Training and Fine-Tuning Environment provides dedicated computational resources for developing, training, and deploying local large language models without relying on external cloud APIs. Built on NVIDIA H100 GPUs with extensive memory capacity, this testbed enables researchers and industry partners to customize open-source language models (Llama, Mistral, BERT variants) for domain-specific applications including technical documentation, customer service automation, and specialized knowledge bases. The environment supports the full LLM lifecycle from pre-training on custom corpora to fine-tuning with techniques like LoRA and QLoRA, and inference optimization for production deployment. This testbed contributes to the CITADELS Framework by democratizing access to state-of-the-art generative AI capabilities while maintaining data sovereignty and privacy.