Hamza B.

I architect production AI video systems: ComfyUI pipelines your backend triggers through an API and that run unattended every day. Two are live now: one inside a profitable real-estate platform, one in beta on a live raffle-draw platform. FEATURED: LIVE RAFFLE-DRAW VIDEO ORCHESTRATION AI lead on the video layer of a live raffle platform. The backend calls the pipeline through the ComfyUI API; it assembles each show from dynamic segments (mascot intro, 3D drum render, winner outro), keeps frame and audio continuity across every cut, adds a configurable interlude with countdown and promo loop, and returns a 1080p stream-ready file through an S3 callback. Delivered milestone by milestone with Swagger and architecture documentation. FEATURED: REAL-ESTATE VIDEO PLATFORM Turns property photos into stable cinematic clips with no manual work. Three stages: GPT Vision routes each photo with a Pydantic schema enforcing the output, the frame is cropped mathematically before generation, then Hailuo 2.3 Pro renders the clip. A scoring engine rates every candidate camera motion for distortion risk before anything is generated. Result: $0.39 to $0.49 per 5-second clip, no jitter or geometry warping, running inside a profitable client platform. WHAT I BUILD • Automated video pipelines: ComfyUI workflows exposed as APIs, multi-segment assembly, encoding for streaming, cloud GPU deployment. • Identity-consistent generation: FLUX character LoRA training, real people kept recognizable across poses and locations, several people in one scene without faces blending. • Computer vision: detection and tracking (YOLOv8, DeepSORT), classification, segmentation and background removal. HOW I WORK • Proof first: a small test on your own material before you commit to the full scope. • Milestones with a pass condition you can check, such as "her face matches the references in 9 of 10 outputs". • A handover your developers can run without me: workflow JSON, API documentation, input specifications and architecture notes. STACK ComfyUI (API, ControlNet, IP-Adapter, custom nodes), FLUX LoRA training, Hailuo, Kling, Seedance 2.0, GPT Vision, Python, FastAPI, YOLOv8, RunPod, DigitalOcean, AWS S3. Send me your current pipeline or your brief. I'll find the constraint that's breaking it and show you a small test before you hire. ### Client feedback “Hamza is a true ComfyUI expert who goes beyond building workflows—he genuinely cares about the project's long-term success. His communication is proactive, clear, and reliable, and his integrity is second to none.” — **Moshe**, Bestiez “Professional and great to work with.” — **Jetnor**