Professional Summary
Computer Vision Engineer and AI Specialist with 2+ years of applied AI experience, based in Islamabad, Pakistan and open to remote or relocation worldwide. Hassan designs and ships end-to-end deep learning pipelines — from dataset preprocessing and keypoint model training to edge and cloud inference — specializing in real-time pose estimation, multi-object tracking and sports analytics.
His production work at ID Sports Ventures in Berlin delivers sub-30ms edge inference, a 15% boost in pose accuracy, and a 40% cut in preprocessing overhead, with 85% test coverage on the systems he ships. He is equally comfortable with C++ and Python, with strong foundations in PyTorch, OpenCV and deployment tooling.
Work Experience
Computer Vision Engineer (Part-Time) · ID Sports Ventures, Berlin, Germany
Jan 2026 — Present
- Architecting scalable real-time video analytics to track athlete performance metrics during live competition sessions.
- Developing lightweight keypoint detection models for edge deployment, achieving sub-30ms latency on streaming footage.
- Collaborating with sports scientists to translate multi-camera pose data into standardized biomechanical reports.
Computer Vision Engineer (Full-Time) · ID Sports Ventures, Berlin, Germany
Sep 2025 — Dec 2025
- Engineered end-to-end CV pipelines for automated athlete motion analysis, boosting pose estimation accuracy by 15% across varied lighting conditions.
- Implemented multi-object tracking (MOT) to handle severe occlusion in dynamic team-sport scenarios.
- Standardized video data workflows, cutting raw video preprocessing overhead by 40%.
- Designed real-time automated scoring and performance assessment engines for training drills.
Software Engineer (Full-Time) · Oxmite Digital Ltd
Jul 2025 — Aug 2025
- Developed backend Python microservices (FastAPI) for high-throughput image and video processing.
- Optimized database queries and containerized core applications with Docker for seamless cloud deployment.
- Raised code coverage to 85% with robust unit-testing suites.
Technical Skills
- Computer Vision — pose estimation, keypoint detection, optical flow, action recognition, object tracking.
- Tracking & Detection — YOLO, ByteTrack, DeepSORT for multi-object tracking through occlusion.
- Frameworks — PyTorch, TensorFlow, OpenCV, TorchScript, ONNX Runtime, TensorRT.
- Programming — Python, C++, CUDA.
- Backend & Ops — FastAPI, Docker, AWS (S3, EC2).
- Streaming — RTSP, FFmpeg for live video pipelines.
- Domains — biomechanics, sports analytics, edge AI.
Featured Projects
AI-Powered Athlete Biomechanics Engine
An automated 2D pose-lifting model that evaluates joint angles and limb velocities during complex athletic movements, using synthetic data augmentation for robustness under high-speed motion blur. Built with PyTorch, OpenCV, ByteTrack, FastAPI and CUDA.
Sub-30ms Edge Pose Inference
Lightweight keypoint detection optimized for edge devices — quantized, TensorRT-accelerated models running on live RTSP streaming footage with minimal latency. Built with TensorRT, ONNX, C++ and FFmpeg.
Multi-Object Tracking for Team Sports
A ByteTrack-based MOT system that maintains consistent athlete identities through severe occlusion, powering automated scoring and drill assessment in real time. Built with YOLO, ByteTrack and Python.
Education
Studied Computer Science, with coursework and self-directed focus on machine learning, image processing and software engineering. Hassan pairs formal foundations with continuous applied learning across deep learning and computer vision.
Contact
- Email — muhammadhassangul01@gmail.com
- LinkedIn — linkedin.com/in/its-muhammad-hassan
- GitHub — github.com/muhammadhassangul01
- Location — Islamabad, Pakistan · Open to remote & relocation.
"I engineer computer vision systems that read human motion at machine speed — from sub-30ms edge inference to production-grade deep learning pipelines." — Muhammad Hassan Gul