class MaiReda:
def __init__(self):
self.role = "Junior AI Engineer"
self.education = "B.Sc. Artificial Intelligence (AOU Egypt + The Open University, UK)"
self.focus = ["Deep Learning", "Computer Vision", "NLP", "Model Evaluation"]
self.learning = ["RAG", "LLM apps", "AI agents"]
self.badge = "Kaggle Notebooks Master"
def what_i_do(self):
return "Data -> Cleaning -> Training -> Benchmarking -> Evaluation -> Deployment"I build deep learning projects end to end: cleaning data, training and comparing models, evaluating them honestly, and shipping the best one as a working app. I'm now moving toward LLM-powered products.
| Stage | What I do |
|---|---|
| 🔍 Explore | EDA, class-balance analysis, data quality checks |
| 🧹 Clean | Duplicate and conflict removal, stratified train/val/test splits |
| 🧠 Model | Custom CNNs, transfer learning (MobileNetV2, EfficientNetB0, DenseNet121, ResNet50, Xception), YOLOv8 |
| 📏 Evaluate | Per-class F1, confusion matrices, ROC-AUC, size vs. speed trade-offs, error analysis |
| 🚀 Ship | Gradio and Streamlit apps |
Cleaned the dataset (13,862 → 3,322 images, 5 classes), then benchmarked a custom CNN against three pretrained models. Deployed with Gradio.
| Model | Accuracy | Macro-F1 | Params |
|---|---|---|---|
| MobileNetV2 | 95.6% | 0.932 | 2.59M |
| DenseNet121 | 94.4% | 0.920 | 7.30M |
| EfficientNetB0 | 94.0% | 0.896 | 4.38M |
| Custom CNN | 68.1% | 0.639 | 1.24M |
TensorFlow/Keras Transfer Learning Gradio
- Deep learning classifier with ROC-AUC evaluation and a tunable decision threshold
- SHAP explanations for each prediction, shown in a Streamlit app
TensorFlow SHAP Streamlit
- Index fingertip as a pen, using MediaPipe Hands (21 landmarks)
- Gesture recognizer for draw, select, erase, and clear, plus undo/redo and shape correction
Python MediaPipe OpenCV
- ANN on 23 network-traffic features: 98.16% accuracy, 98.75% F1, 99.70% ROC-AUC
- Streamlit app for batch prediction from CSV files
TensorFlow Scikit-learn Streamlit
More machine learning, deep learning, NLP, computer vision, and data analysis projects.
🧠 Languages & data
📊 Analysis & visualization
🤖 Machine learning & deep learning
🚀 Deployment & tools
| 🥇 | Kaggle Notebooks Master (profile) |
| 🏅 | First Place, Team Project (Iron Engineer), Microsoft Club, AOU Egypt, 2024 |
| 🎖️ | Honorable Mention, ICPC ECPC 2024 Qualifications, Alexandria |
- B.Sc. in Artificial Intelligence, Arab Open University – Egypt Branch (dual degree with The Open University, UK), 2021 – 2025
- AI Diploma (170 hours), Instant Software Solutions
- Docker for the Absolute Beginner (Packt) · Get Started with Jira (Coursera Project Network) · Google Cloud Skills Boost: Cassandra on Google Cloud
- Learning RAG, vector databases, and LLM agents
- Improving my models with cleaner data splits and stronger evaluation
Open to Junior AI Engineer roles, remote or on-site in Cairo.
⭐ Thanks for stopping by!