AI Infrastructure Engineer

Role overview Design and scale the GPU inference infrastructure behind an image-generation product. The focus is on replacing workflow-based processing with a stable production architecture that runs models efficiently, handles asynchronous generation jobs, and supports multiple model workflows.

Responsibilities - Design and optimize an inference pipeline using Hugging Face Diffusers and PyTorch. - Deploy and manage containerized GPU environments with Docker, CUDA, and PyTorch. - Reduce model startup time and manage GPU memory use, including VRAM constraints. - Build scalable asynchronous job queues for image-generation requests. - Improve inference speed, reliability, and operating cost. - Support multiple model workflows and contribute to longer-term infrastructure decisions.

Requirements - Strong hands-on experience with PyTorch and Hugging Face Diffusers. - Deep understanding of CUDA and GPU memory management. - Experience deploying machine-learning models in Docker containers. - Experience building scalable backend inference services. - Strong debugging and performance optimization skills.

Nice to have - Experience with SDXL, Flux, or LoRA implementations. - Experience with GPU cloud providers such as RunPod or AWS. - Experience optimizing inference pipelines for production.

Benefits and work setup - Full-time freelance collaboration, remote, with substantial ownership and autonomy. - Long-term collaboration is intended.