Ahmed Abdelgelel Zahran

Machine Learning Engineer · Cairo, Egypt

I build end-to-end ML systems — from model training and evaluation to cloud deployment and real-world integration. My background spans computer vision, applied deep learning, and AWS MLOps. I care about shipping things that actually work, not just notebooks that run locally.

Currently building: MED Query — a medical RAG agent with FastAPI backend.

---

What I Work With

ML & Deep Learning — PyTorch · scikit-learn · EfficientNet · YOLOv8 · CatBoost · Hugging Face
MLOps & Cloud — AWS SageMaker · Lambda · Step Functions · MLflow · Docker · GitHub Actions
Agentic AI — LangChain · ChromaDB · FastAPI · RAG pipelines
Data — Pandas · NumPy · Matplotlib · SQL
Languages — Python · C++

---

Projects

🔬 MED Query — Medical RAG Agent (in progress)

Medical Q&A agent that retrieves answers from clinical guidelines and drug documentation, citing exact sources. Built with LangChain + ChromaDB + Groq + FastAPI.

🎯 Landmark Classification — Live Demo

Benchmarked 3 CNN architectures (scratch · residual · ResNet34) on 50 world landmark classes. Best model: 74.8% F1. Includes Grad-CAM explainability and a live Streamlit comparison app.

🔐 Security Detection System

Real-time weapon detection + face recognition pipeline using YOLOv8 and LBPH. Three-case threat logic (unknown intruder · restricted user · known criminal) with auto-alarm and screenshot capture. +88% F1-score. → Watch demo

🩺 Skin Cancer Detection

Multimodal classifier combining EfficientNet (lesion images) + CatBoost (patient metadata) on the ISIC 2024 dataset — 401,000+ samples. Deployed via AWS Lambda + API Gateway.

☁️ ML Workflow on AWS SageMaker

Event-driven image classification pipeline: SageMaker → Lambda → Step Functions with confidence-gated routing and MLflow experiment tracking.

🚲 Bike Sharing Demand Prediction

Tabular forecasting with AutoGluon on AWS SageMaker. EDA-driven feature selection + hyperparameter tuning to optimize RMSE.

🛠️ Freelancing — Automation & Data

Web scraping (BeautifulSoup), form automation (Selenium), and data manipulation pipelines (Pandas) built for client deliverables.

🌐 Network Projects ⭐5

CCNA/CCNP hackathon tasks: multi-branch VLANs, OSPF/EIGRP redistribution, BGP/MPLS inter-AS routing, Dynamic PAT. Built at Orange Digital Center.

---

Open Source

Unify AI / ivy-llc — Integrated jax.lax.scan into the Ivy cross-framework ML library. Developed unit test suites for PyTorch, JAX, and TensorFlow backends. (14.2K+ ⭐) PR #22412

---

Certifications

AWS Machine Learning Engineer Nanodegree — Udacity & AWS · 2024

---