r
radwahashiesh

Radwa

@radwahashiesh

Computer Vision Medical AI Engineer Deep Learning Specialist

Égypte
Arabe, Anglais
Certaines informations sont présentées en anglais.
À propos de moi
Machine Learning Engineer specializing in Computer Vision, Deep Learning, and Medical Image Analysis. I have hands-on experience building advanced AI systems including medical image segmentation using Transformer-based architectures, hyperspectral image classification, and machine learning optimization models. My expertise includes Computer Vision (Segmentation, Classification, Detection) Deep Learning model development (PyTorch / TensorFlow) Data preprocessing & feature engineering Model optimization and performance improvement I have successfully completed 40+ projects with 5-star feedback ... Plus d’infos

Compétences

r
radwahashiesh
Radwa
hors ligne • 
Temps de réponse moyen de 1 heure

Voir mes services

Vision par ordinateur
I will build a medical image segmentation model using deep learning
De90 $US / projetPlus de détails

Portfolio

Expérience professionnelle

Upwork

Computer Vision Engineer | Medical Imaging Deep Learning Built cross-modal segmentation using Swin Transformer for multi-modality generalization. Objectives: Generalization across modalities, improving imbalanced segmentation, stable efficient training. Solution: Custom SwinSegAblationCustom with Hybrid Cross Attention, FFT Dynamic Frequency Analyzer, Multi-Head Attention in modular ablation-ready design. Training: BCEWithLogits + Dice, AdamW lr 1e-4 wd 1e-5, cosine annealing 25 epochs, AMP, gradient clipping. Data: 80/20 split, 224×224 resizing, flips/rotations/brightness/contrast/saturation augmentations. Metrics: Dice, IoU, Precision, Recall, Cohen’s Kappa (accuracy not used due to imbalance). Improvements: Stabilized FFT (float32, no autocast), safer GPU ops, gradient clipping, stronger augmentation. Results: Improved Dice/IoU with stable convergence. Innovation: Combines spatial + frequency learning with modular experimentation and efficient training. Impact: Scalable cross-modal segmentation for research and clinical use. Stack: Python, PyTorch, Swin Transformer | 5.0 client rating.

Upwork • Freelance

Dec 2021 - Dec 2025 • 4 yrs

skilled at machine learning and deep learning specialization, build model with different algorithms and visualize the results