Researchers have developed an end-to-end adaptation of NVIDIA's Nemotron retrieval stack specifically for Modern Greek, addressing the language's absence in major multilingual retrieval benchmarks. The methodology involved extensive corpus mining and synthetic supervision to facilitate retrieval-augmented generation (RAG) across specialized domains like law, medicine, finance, and energy. By fine-tuning a Nemotron 1B embedder on 65,773 Greek retrieval pairs, the authors achieved a significant increase in nDCG@10, rising from 0.362 to 0.835, which notably outperformed off-the-shelf multilingual dense retrieval models and even the BM25 baseline in certain specialist contexts. The adaptation extended to a cross-encoder reranker and a LoRA-tuned Nemotron 30B-A3B mixture-of-experts (MoE) reader. This fine-tuning for grounded generation boosted judged answer correctness from 29.4% to 66.9%, while simultaneously enhancing citation quality and faithfulness. To support future research, the study introduces HERA, the first large-scale Greek benchmark designed specifically for evaluating RAG systems. The resulting models and benchmark have been released to mitigate the current lack of high-quality Greek-language RAG resources in specialist domains.