Senior Backend & GPU Infrastructure Engineer – Headless ComfyUI / AI Video Pipeline

We are building an AI Video Generation SaaS Platform, and are looking for an experienced Backend & GPU Infrastructure Specialist to design and deploy our cloud video generation backend pipeline. We plan to self-host high-performance open-source video models (such as LTX-Video, MiniMax Hailuo / Wan) on cloud GPU platforms (RunPod, Vast.ai, Modal), using ComfyUI in headless/API mode integrated into our core API backend. Key Responsibilities ComfyUI Pipeline Automation: Convert video generation workflows into headless ComfyUI API blueprints (prompt JSON execution) that run programmatically without UI overhead. GPU Cloud Infrastructure & Deployment: Deploy, containerize (Docker/CUDA), and manage GPU instances on providers like RunPod (Serverless / Network Volumes), Vast.ai, or Modal. Async Queue & Job Management: Build a scalable, fault-tolerant job queue (Redis + Celery / BullMQ / FastAPI) to queue incoming user video requests, track render progress, and deliver Webhook/WebSocket updates to our web frontend. Storage & CDN Integration: Set up fast, direct uploads of finished video outputs to cloud storage (S3 / Cloudflare R2 / DigitalOcean Spaces) with signed CDN links. Autoscaling & Cost Optimization: Implement cold-start optimization and auto-scaling logic to keep operational GPU costs low during low-traffic periods while handling spikes smoothly. Required Experience & Tech Stack AI Media Pipelines: Proven experience running headless ComfyUI API workflows for image/video generation at scale. GPU Cloud Providers: Hands-on deployment experience with RunPod (Serverless or Pods), Vast.ai, Modal, or Salad. Backend Languages: Strong proficiency in Python (FastAPI / Celery) or Node.js/TypeScript. Infrastructure & Storage: Docker, CUDA drivers, PyTorch model deployment, Cloudflare R2 / AWS S3. Asynchronous Architecture: Redis, Webhooks, or WebSockets for live job status reporting. Preferred Qualifications Prior experience building commercial AI SaaS media generation platforms (image, voice, or video).