I’m looking for an experienced Linux/Python/ai engineer to provision and configure a gpu server on
runpod.io and build a fully automated video VFX pipeline:
Infrastructure Setup on
RunPod.io
OS Image: Ubuntu (ubuntu:latest)
Hardware: 1–2 × H100 PCIe GPUs, 2 vCPUs
Persistent Storage: 50 GB network volume (no data loss on restart)
Billing: On-demand (all infrastructure costs covered by client)
Automated Installation Script (Bash)
Install Blender 4.x with CUDA/Mantaflow support
Create and activate a Python 3 venv, install ComfyUI
Clone and install custom ComfyUI nodes:
Segment Anything (SAM)
XMem
Video I/O node
Fluid Mask node
Generate two executable scripts:
run_pipeline.sh: runs SAM → XMem tracking → fluid simulation in Blender → composite final video
process_video.sh: one-line CLI for user upload & run by specifying input file and time window
User-Friendly Web Dashboard (Flask)
Video upload interface
“Start (s)” / “End (s)” input fields
Preview gallery showing frames every X seconds
“Generate” button to kick off the pipeline
“Download” link for the finished MP4
Testing & Documentation
Validate with an HD (720×1080) sample video
Provide clear instructions for use, maintenance, and redeployment
Skills & Qualifications:
Proficient with Ubuntu/Debian and Bash scripting
Strong Python background, virtualenv, pip/package management
Experience with Flask (or equivalent) for building lightweight dashboards
Hands-on with Blender cli, cuda, mantaflow for fluid vfx
familiarity with comfyui and creating custom python nodes
knowledge of sam + xmem for accurate video object tracking
git/version control and devops best practices
what’s provided:
runpod.io access with H100 nodes and 50 GB persistent volume
Full administrative privileges (no infra costs to you)
Freedom to choose Docker or plain-VM approach
Delivery term: Not specified