Semir (森马)

Solution

Semir adopted Alibaba Cloud Function Compute as the serverless base for its AIGC design workflow and used Stable Diffusion as the image- generation engine. Function Compute provides virtualized GPU resources that can be scaled per request instead of through long-lived cluster planning, and it supports multiple card types plus very small 1 GB GPU slices so design workloads can start cheaply and scale only when needed. Around the model itself, Semir built AI inference applications and serving logic on top of the serverless runtime so designers could submit prompts, switch models without heavy tenant interference, and avoid the long queues common in shared local WebUI setups. Because the platform includes monitoring, logging, and alarm capabilities by default, the company could focus on testing how AIGC fits apparel design instead of first building a complex ops stack. The result was a practical, low-friction path from experimentation to production design support.

Data flow

Designers submit prompts into the Function Compute-based inference application, where Stable Diffusion loads on elastic GPU resources, generates candidate design drafts, and returns them to the apparel workflow. Monitoring and scaling are handled by the serverless platform so peak creative demand can be absorbed without a dedicated cluster-operations team.