Machine Learning Engineer. Computer Vision Specialist. Science First Mindset.
I build intelligent systems that see, understand, and react to the world. My work blends deep learning, computer vision, and large-scale ML engineering to solve high-impact real-world problems.
###What I Am Learning
I regularly explore new ideas in deep learning, computer vision, and optimization. I enjoy reading research papers and turning theoretical concepts into working prototypes. Deep Learning TA for more than 80 students at LUMS. Helped shape assignments, quizzes, and tutorials.
I enjoy taking ideas from research and turning them into robust pipelines that run on real hardware, real users, and real constraints.
🧩 Seven years of hands-on engineering
📷 Experience across 2D, 3D, temporal and multimodal ML
🎓 Research background from Computer Vision and Graphics Lab (CVGL)
🔍 Strong focus on reproducibility and clean engineering
🌍 Collaborations with WWF and IWMB on wildlife conservation
🐌 I support slow science
- Driver monitoring and safety features
- Object detection, segmentation, and classification
- Multi-camera surveillance and tracking
- Model optimizations for inference on edge devices
- Graph Neural Networks for remote sensing
- Spatio-temporal modeling
- NeRF feasibility studies
- Representation learning for low data settings
- Distributed training on EC2
- Dataset pipelines at scale
- Real-world deployment for production
- Automated data ingestion, ETL, and monitoring
Models for distraction, mobile phone usage, drowsiness, and unsafe behaviors.
Real-time performance with unified multi-task architecture.
Published work in the AAAI 2023 Fall Symposium.
Designed T RAG to encode temporal information through region graphs.
IoT-powered camera traps, real-time inference, solar-powered nodes, Django backends, and alerting systems for remote habitats.
GNNs trained on 3D point clouds reconstructed from 2D imagery.
Evaluated NeRFs for industrial use cases.
Multi-camera geometry fused with YOLO-based detection for tracking and spatial reasoning.
Spatio Temporal Driven Attention Graph Neural Network with Block Adjacency Matrix (STAG NN BA)
Authors: U. Nazir, W. Islam, and M. Taj
The work proposed a spatiotemporal attention-based graph neural network model designed for land-use change detection across temporal satellite imagery. It was presented at the AAAI 2023 Fall Symposium Series.
| Project | Description | Tech |
|---|---|---|
| driver-monitoring | Unified object detection and behavior classification for road safety | PyTorch, CV, Edge Inference |
| remote sensing gnn | Temporal GNN for land use change detection | GNNs, PyTorch Geometric |
| wildlife early warning | IoT trap system for endangered species | Jetson Nano, Django, EfficientNet |
| nerf research | Experiments on NeRF for structured scenes | NeRF, 3D Learning |
| surveillance mvp | Multi camera CV surveillance MVP | YOLO, Geometry |
- Multimodal perception for safety systems
- Self-supervised learning for vision
- Spatio-temporal modeling
- 3D scene reconstruction
- Efficient architectures for edge devices
- Reliability, calibration, and model trustworthiness
Languages: Python, C++, Javascript, SQL
Frameworks: PyTorch, TensorFlow, Django, React
Vision Models: YOLO v3 and v5, EfficientNet, U-Net, MegaDetector
GNN Models: GCN, GATv2, MoNet, SAGE-Net
3D and Novel View: GroupFree3D, Photogrammetry, Structure From Motion
Infra: AWS EC2, Docker, Scrapy, Jetson Nano
Specialties: Data pipelines, edge inference, geometric learning, multimodal learning, perception
I love teaming up for:
- Applied deep learning research
- Open source CV tools
- Vision, language, and multimodal projects
- Scientific implementations
- Wildlife conservation tech
- Real-world detection and tracking systems
Email: wadoodislam@gmail.com
LinkedIn: linkedin.com/in/wadood-islam
✨ Thank you for visiting. Feel free to explore, star anything you find helpful, or reach out for collaboration.
