Based solely on the provided article, foundation models like ESM2 and Evo 2 are reshaping computational biology. Adapting these billion-parameter models to specific tasks is nontrivial, as full fine-tuning is impractical. Low-Rank Adaptation (LoRA) addresses this by keeping the pretrained backbone frozen and training only a small set of low-rank adapter matrices, training ~1% of the parameters and fitting a single billion-scale model on a single workstation GPU. NVIDIA BioNeMo Recipes provide step-by-step training recipes built on PyTorch, Hugging Face, and Megatron-Bridge patterns. The post walks through two case studies on a single NVIDIA RTX 6000 Blackwell Workstation Edition GPU: ESM2-3B plus LoRA for protein secondary structure prediction (PSSP) and Evo2-1B plus LoRA for DNA splice-site classification. All source code is available in NVIDIA BioNeMo Recipes.