Computer Vision & Robotics Software Engineer
HiWi Student Research Assistant at the Humanoid Robots Lab, University of Bonn. Focused on active perception, intelligent robot manipulation, and embodied AI.
About Me
Results-driven Computer Vision Software Engineer with 2+ years of experience in robotics and real-time video analytics. I currently serve as a HiWi Student Research Assistant at the Humanoid Robots Lab (University of Bonn), focusing on active perception and intelligent robot manipulation.
I build production-level C++/Python microservices and deep learning pipelines that improve performance and accuracy. My technical focus spans 3D perception, camera calibration, photogrammetry, point clouds, COLMAP, NeRFs, Gaussian Splatting, and CLIP.
I enjoy translating research into reliable systems for embodied AI and I am open to collaborations in robot perception, learning, and autonomous systems.
Highlights & Achievements
Certifications
- Fundamentals of Reinforcement Learning (Coursera, University of Alberta)
- Introduction to Intel OpenVino (Coursera, Intel)
- Robot Perception (Coursera, University of Pennsylvania)
- Neural Networks and Deep Learning (Coursera, Andrew Ng)
- Transformer and BERT Model (Coursera, Google)
- Computer Vision Basics (Coursera, University of Buffalo)
- Master SolidWorks
Research Focus
- Active Perception
- Intelligent Robot Manipulation
- Embodied AI & Robot Learning
Language Skills
- English (C1)
- French (B1)
- German (A1)
Publications
- R. A. Gull, M. D. Bin Mohamed Izham and J. Qadir, "1 Robotics Primer for Independent Learners: Background, Curriculum, Resources, and Tips," 2023 IEEE Global Engineering Education Conference (EDUCON), Kuwait, Kuwait, 2023, pp. 1-9, doi: 10.1109/EDUCON54358.2023.10125209. PDF
Technical Skills
Computer Vision
- OpenCV & Open3D
- Object Detection & Tracking
- Visual Perception
- Monocular Depth Estimation
- Point Clouds & COLMAP
- Camera Calibration
- Photogrammetry & 3D Reconstruction
- NeRFs & Gaussian Splatting
- CLIP
Data Analytics & Utilities
- Seaborn
- Pandas
- Numpy
- Eigen
- Ceres
- JSON
- MQTT
Machine Learning & Deep Learning
- PyTorch
- LSTM
- UNet
- ResNet
- RetinaNet
- CNN
Programming Languages & Scripting
- Python3
- C++
- Rust
- CUDA
- Java
- Matlab
- Bash Scripting
Robotics
- ROS2
- CoppeliaSim
- Webots
- IsaacSim & IsaacLab
- Teleoperation
Software Development
- Microservices
- FastAPI
- RestAPI
- Protobuffer
Tools & Frameworks
- Git
- Docker
- CI/CD
- Unit Testing
Databases
- Redis
- SQLite
Professional Experience
Student Research Assistant (HiWi)
Humanoid Robots Lab, Universität Bonn, Bonn, Deutschland
- Active perception and intelligent robot manipulation.
Computer Vision Engineer Intern
Horizon Telecom, Montceau-les-Mines, France
- Deployed the system from simulation to real-world Sim2Real, integrating it with Doosan M-series Robot controls.
- Developed a 3D scene understanding perception system for electronics warehousing.
Software Engineer
HAZEN.AI (Remote), Mecca, Saudi Arabia
- Optimized real-time video analytics by developing CUDA C++ kernels for NVIDIA Jetson devices.
- Added MQTT and Redis-based service to pub-sub tarfiles, images, and path of a folder using Google protobuf.
- Contributed in end-to-end python application development, including launching from scratch, within an agile team of 7 for AI for road safety applications.
- Developed cross-platform containerized services using Docker for arm64 and x86 architectures, streamlining deployment workflows.
- Managed global deployments across 9+ sites and ensured python code quality with 92% test coverage.
- Accelerated image feature correspondence by 600% and improved vehicle speed prediction accuracy by optimizing GPU-accelerated distortion correction and object segmentation within our 3D reconstruction and structure-from-motion pipeline.
Deep Learning Research Intern
BioRobotics Institute (Scuola Superiore Sant’Anna), Pisa, Italy (Remote)
- Deployed LSTM model to an STM32 IoT Node for real-time edge inference, enabling on-device predictions.
- Engineered ML preprocessing pipelines for time-series data, improving data quality by 23%.
- Achieved 97.4% prediction accuracy by developing and refining a Python LSTM model.
Projects
Mimic Me Robot
Designed an algorithm and coded a UR-5 robotic manipulator in a simulated environment to use an RGB camera and mimic human arm motion in real time. OpenCV | Robot teleoperation | Mediapipe | Webots | Robot simulation.
VP Subject Classification
Developed a Kohonen Self-Organizing Map (SOM) for classification and identification of human subjects from marker data, and a biologically plausible SNN using Spike-Timing-Dependent Plasticity (STDP) to learn spatiotemporal patterns. Spiking Neural Networks (SNNs) | STDP | Python | NumPy.
MLOps Deployments on Cerebrium Serverless Platform
Converted a pre-trained PyTorch image classification model to ONNX with embedded preprocessing, built Python wrappers for ONNX Runtime, and deployed a containerized service on Cerebrium with auto-scaling. Integrated FastAPI endpoints and achieved a 1.2s E2E response time (inference + API latency). Python | FastAPI | PyTorch | ONNX | Docker | Serverless Deployment | MLOps | Cerebrium.
HBP-Measurements
Engineered a perception system to estimate wrist and body measurements using a single RGB image and height input. Leveraged SMPL and deep learning to infer body dimensions in centimeters, delivering real-time performance at 1 FPS. OpenCV | SMPL | Human Pose Estimation | 3D Digital Humans | TensorFlow | Pillow | NumPy | Qt.
Incremental Structure From Motion SFM
Designed and implemented an incremental SfM pipeline from scratch in Python, reconstructing 3D models from unstructured images. Performed camera calibration for mobile phones and Intel RealSense cameras, extracted SIFT features, and matched correspondences with OpenCV BFMatcher. Optimized bundle adjustment with SciPy least-squares and PyTorch Adam, achieving 4x faster convergence in reprojection error minimization. Added point cloud colorization and exported cameras, images, and 3D points to COLMAP for validation. OpenCV | PyTorch | COLMAP | 3D Reconstruction.
3D Object Detection RetinaNet for Autonomous Robotics
Optimized an open-source 3D-RetinaNet model with ResNet50 backbone + FPN for real-time 3D scene understanding, achieving 87% pedestrian detection accuracy. Enhanced inference speed to 25 FPS for autonomous robot video feeds and built an annotation preprocessing script to streamline 3D bounding box training. 3D-RetinaNet | ResNet50 | Feature Pyramid Network | Real-Time Inference.
Education
Master of Science in Mobile Robotics
University of Bonn, Bonn, Deutschland
October 2025 - Present. Focused on autonomous systems, computer vision, sensing, and machine learning for intelligent robotic systems.
VIBOT M1 Master of Science in Computer Vision
Université de Bourgogne Europe, Le Creusot, France
September 2024 - June 2025. Focused on robot perception and 3D computer vision.
Bachelor of Science in Electrical Engineering
Information Technology University, Lahore, Pakistan
September 2019 - May 2023. Focused on Computer Engineering.
Get In Touch
I'm open to research collaborations and engineering roles in robot perception, learning, and embodied AI.