Self-Hosted AI Production Server and Automated Content Pipeline
Built and operate a 24/7 AI production server on Ubuntu 24.04 running Ollama, ComfyUI on an RTX 4070, Whisper, and n8n across a custom Docker network. Solved cross-container networking, OAuth2 integration, and FFmpeg codec issues while building NOC-grade monitoring and a fully automated video content pipeline.
A 24/7 self-hosted AI production server running 12+ Dockerised services including LLM inference, image generation, audio transcription, and workflow automation. Built end-to-end automated video content pipelines that generate, assemble, and publish videos without any manual steps per cycle.
.outcomes
01GPU passthrough configured for real-time image generation via ComfyUI on RTX 4070
02Custom Docker networking with static IP, SSH hardening, firewall rules, and DNS routing
03FFmpeg pipeline handling multi-format output (9:16 and 16:9) with cross-service audio codec resolution
#Docker#GPU Computing#n8n#FFmpeg#Linux#Self-Hosted AI
.gallery
Generative AI imagery — Flux Schnell diffusion model. Prompt: Modern mountain cabin at dusk, floor-to-ceiling glass walls, warm interior light spilling onto snow, pine forest backdrop, photorealistic architectural render, 8k
Generative AI imagery — Flux Schnell diffusion model. Prompt: Close-up portrait of an astronaut inside a capsule, helmet reflection showing Earth, dramatic rim lighting, shallow depth of field, ultra-detailed, cinematic
Generative AI imagery — Flux Schnell diffusion model. Prompt: Overgrown brutalist library, vines reclaiming concrete shelves, shafts of light through broken skylights, post-apocalyptic, concept art, cinematic atmosphere
