Fashion AI Model Optimization Expert

I’m building a fashion-centric generative AI and now need a specialist who can take our LoRA-based SDXL pipelines and push them to production quality. The core focus is model training and optimization: from curating and augmenting our annotated apparel dataset, through precise fine-tuning of diffusion checkpoints, to rigorous performance testing so we know each release actually improves style fidelity and fit accuracy.

Everything currently lives in a PyTorch environment with Diffusers and the latest SDXL weights on Hugging Face. You’ll work directly inside that stack—no need to port anything—but I’m expecting you to bring best-practice tricks for LoRA composition, checkpoint merging, and custom schedulers that shave both VRAM and training time.

Deliverables I need from you
• A reproducible training pipeline (scripts + README) that covers data preprocessing, LoRA injection, and scheduled training runs
• At least one fine-tuned SDXL checkpoint ready for fashion look-book generation, along with quantitative and visual evaluation reports
• A brief performance analysis outlining what worked, what didn’t, and clear next steps I can hand to the next engineer or pick up myself

If you’ve already wrangled diffusion models for e-commerce, virtual try-ons, or style transfer, you’ll feel right at home here. Let me know how soon you can start and point me to any fashion-focused samples you’ve shipped so I can see your touch in action.