Nadhir H.

Design Future Ready AI Vision Systems That Deliver Real Results Machine Learning and Computer Vision engineer with nearly 7 years of hands-on experience. I take on complex vision and AI problems and turn them into working systems, from first prototype through to production deployment. Recent Recognition 🏆 1st Place, The FREUID Challenge 2026 at IJCAI-ECAI. Won an international AI competition held as part of IJCAI-ECAI 2026, and was invited by the organizers to present the winning solution at the conference in Bremen, Germany. IJCAI has been running since 1969 and is one of the most established AI research conferences in the world. 🥈 Kaggle Silver Medal. Top 100 out of more than 5,000 participants in the Orbit Wars reinforcement learning competition. 🏅 1st Place, Machine Learning Hackathon. Certificate available in my portfolio. 💯 100% job success rate across completed client contracts. I previously held Top Rated and Top Rated Plus status. I stepped back from new contracts for roughly a year while committed to several long running projects, and the badges lapsed from inactivity rather than performance. My full ratings history is visible on this profile. What I Build Detection, tracking and analytics: Real time vehicle detection with speed estimation and traffic monitoring. Shop footfall counting. Soccer match object detection, player tracking and activity recognition. Basketball training analysis covering player form and shot trajectory. Strawberry detection and yield estimation for agriculture. Recognition and classification: Fastener recognition identifying screws, bolts and washers and mapping them to part numbers, built as a full detection, segmentation and classification pipeline. Face recognition attendance systems. Image classification with CNNs and Vision Transformers. Real time ID card information extraction. Automated MCQ grading. Pose, gesture and 3D: Full body pose estimation and gesture classification. Computer vision games for kids. Hand gesture volume control. 3D depth camera work including point cloud capture and visualization. NLP and conversational AI: End to end IVR systems combining speech recognition with language understanding. Conversational chatbots with intent recognition, context handling and custom actions. LLM integration and prompting. Other: Pool shot outcome prediction. Green screen removal. Deepfake research pipelines. Technical Stack Languages: Python (primary), C++ Deep learning: PyTorch, Keras, Torchvision, fastai, Hugging Face Transformers, ONNX Architectures: YOLO (v3 through v11 and NAS), Faster R-CNN, Mask R-CNN, RetinaNet, SSD, U-Net, EfficientNet, ResNet, Vision Transformers, CNNs, RNNs, LSTMs Computer vision: OpenCV, MMDetection, Detectron2, Albumentations, MediaPipe, DeepFace, face_recognition, Segmentation Models PyTorch, Tesseract OCR, PaddleOCR, EasyOCR Data and classical ML: scikit-learn, pandas, NumPy, SciPy, Matplotlib, Seaborn, Plotly, H2O.ai, AutoML Deployment: FastAPI, Django, Streamlit, Tkinter, AWS EC2, Google Cloud, Azure, Windows, Linux, macOS How I Work I own projects end to end. That means data collection and annotation, augmentation pipelines, model selection and fine tuning, evaluation that goes beyond headline accuracy, and deployment behind an API the client can actually use. I fine tune strong pretrained models rather than building from scratch, because on real client datasets that consistently gives better accuracy in less time. I also write plain language updates throughout, so you always know what moved, what did not, and what comes next. To me, earning money is way less important than building trust.