NKD Article Image Generation

Fine-tuned Stable Diffusion XL using LoRA on brand-specific image datasets; built a Python pipeline for structured prompt construction, batch image generation, and CLIP-based cosine similarity quality filtering to discard off-brand outputs. Delivered a reusable GenAI pipeline that significantly reduced production time and replaced manual photography with consistent, on-brand, production-ready article imagery.

NKD's product teams relied on slow manual photography and inconsistent stock visuals for article imagery, creating bottlenecks and brand inconsistency at scale.

SDXL + LoRA fine-tuned on brand-specific datasets with CLIP-based quality filtering pipeline for scalable, on-brand article image generation.