Build a High-Quality AI Image Generation Inference Pipeline

We're looking for an experienced ML/MLOps engineer to design and deploy an inference pipeline that generates high-quality AI images from text prompts. The goal is a reliable, production-ready system we can call via API to produce images at consistent quality, with reasonable latency and cost.

Scope of Work

- Select and deploy a state-of-the-art text-to-image model (e.g., FLUX, Stable Diffusion 3.5 / SDXL, or your recommendation based on our needs)

- Build a clean inference API (REST endpoint) that accepts a prompt and parameters and returns generated images

- Optimize output quality: schedulers/samplers, refiners, and upscaling where appropriate

- Set up GPU-backed deployment

- Optimize for latency, throughput, and cost; implement queueing/batching if needed

- Provide documentation covering setup, API usage, parameters, and how to update or swap models

Required Skills

- Strong experience with diffusion models and image-generation inference

- Python + PyTorch; familiarity with Hugging Face Diffusers

- GPU deployment and serving experience (containers, autoscaling, cold-start handling)

- API development and basic security (auth, rate limiting)

Nice to Have

- Experience with LoRA, ControlNet, IP-Adapter, or fine-tuning

- Prompt engineering and quality benchmarking

- Cost-optimization on serverless GPU platforms

Deliverables

- Deployed, documented inference endpoint

- Source code in a repository we own

- Brief handoff doc or call