BS Software Engineering
COMSATS University Islamabad
2021 – 2025
Final year project: Hoop Master, an AI-assisted basketball training application. I owned the computer-vision backend.
Hello, I’m
Huzaifa Qadeer
Full-Stack AI Engineer
I build AI systems for video. My work covers computer vision, generation pipelines and the APIs behind them, mostly in Python. I have been shipping these systems to production for YouTube channels since 2025.
About
Education and practice
I finished my BS in Software Engineering at COMSATS University Islamabad in July 2025. Since August I have worked independently with a small group of collaborators, building and running these systems for YouTube channels.
Education
Engineering practice
AI Developer
Independent, with a small group of collaborators
Aug 2025 – Present
- Built the pipeline that takes a script to a finished video: storyboard, voiceover, timing alignment, footage acquisition, motion graphics, render.
- Split asset acquisition and motion design into standalone services, Source IT and the Motion Graphics API, so more than one product can call them.
- Operate the systems in production for YouTube channels.
Skills
Skills and stack
Languages
Python first, TypeScript for anything that renders.
- Pythonprimary
- TypeScript
- JavaScript
AI & computer vision
Pose estimation, video understanding, and model serving.
- PyTorch
- OpenCV
- MediaPipe Pose
- LLM orchestration
- Voice synthesis
Video & rendering
The layer that actually produces the finished file.
- FFmpeg
- Remotion
- Hyper Frames
- Motion graphics
Backend & APIs
Services that other services call.
- Flask
- Node.js
- Express
- REST API design
- Queueing
Data & infrastructure
Where things are stored and where they run.
- MongoDB
- AWS
- RunPod
- GPU inference
Frontend
Interfaces for the tools, and one mobile app.
- React
- React NativeFYP
- Next.js
Tools
Day to day.
- Git
- GitHub
- VS Code
- Figma
- Premiere Pro
Work
Projects
- FinishedNot hosted
MotionGen AI
Script in, finished video out, with a scene review step before the final stitch.
- Motion Graphics API
- Source IT
- Remotion
- FinishedHosted
Source IT
One API for footage and asset sourcing across licensed and public-domain libraries.
- HTTP API
- AWS
- FinishedNot hosted
Motion Graphics API
80+ hand-built motion-graphics components, tested one by one and served behind an API.
- Remotion
- React
- TypeScript
- HTTP API
- UnfinishedNot hosted
Documentary Studio
Fully automated video production for formats that do not need a review pass.
- FFmpeg
- Hyper Frames
- Built-in SERP lookup
- FinishedNot hosted
Hoop Master
AI-assisted basketball training — pose-based analysis of shooting and drill form.
- Python
- MediaPipe Pose
- OpenCV
- PyTorch
- UnfinishedNot hosted
Self-Hosted Video Generation
Running open video-generation models on infrastructure I control.
- RunPod
- B200 GPUs
- FP8 quantization
System
How it fits together
- 01
Acquisition
Source IT
- Unified search across licensed and public-domain libraries
- One normalized response shape, attribution metadata per clip
- Queueing, deduplication, partial fetching
- 02
Render
Motion Graphics API
- 80+ hand-built Remotion components
- Maps, typography, parallax, diagrams, charts, camera movement
- Served over an API, so no single product owns it
- 03
Orchestration
MotionGen AI
- Calls both services to take a script to a finished video
- Human in the loop: scene-level regeneration before the final stitch
- The only product that composes the two services
- 04
Runs on its own
Documentary Studio
- Fully automated, template-driven, no review gate
- FFmpeg assembly with Hyper Frames, plus its own narrow sourcing lookup
- Shares no services with the stack above
Contact
Get in touch
Every repository here is private. Ask for a walkthrough of the architecture or of any part of the code.
huzaifaqadeer13@gmail.com