Generative AI — Model Training & Visual Studies by Wenyu XiaGenerative AI — Model Training & Visual Studies by Wenyu Xia

Generative AI — Model Training & Visual Studies

Portfolio project
Wenyu Xia
Role: Product Manager & Model Training
Generative-AI models and image workflows for architecture, interiors, products and urban design.
Work delivered • Dataset preparation. • LoRA and checkpoint training. • Model testing. • Reusable workflow documentation.
Overview: Training experience and visual concept studies are presented separately. The studies below do not establish model-training results or client deployment.

Architecture light study

A self-initiated study explores daylight and blue-hour appearances around an architectural reference. It is a visual concept exercise, not a claim of authorship of a constructed building, an engineering simulation or proof of a trained model.
Architecture light study — AI-assisted visual concept; daylight and blue-hour comparison.
Architecture light study — AI-assisted visual concept; daylight and blue-hour comparison.

Product visual study

A fictional jewelry concept explores reference-connected imagery and recognizable product details. Results require visual review and do not demonstrate exact reproduction of a real product SKU.
Fictional jewelry visual study — AI-generated concept image; not a real product photograph.
Fictional jewelry visual study — AI-generated concept image; not a real product photograph.

0 likes · 1 view · Oct 5, 2026

Model-training experience, architecture light studies and fictional product imagery, with training work separated from visual concept experiments.

Tags