Junior Teaching Assistant
German University in Cairo
Taught Data Structures and Algorithms in Java, guiding students through worksheets and explaining core concepts in lab sessions.
Hi, I am
Youssef Magdy
>AI & Machine Learning Engineer
I build intelligent systems — from Transformers written from scratch in PyTorch to retrieval-augmented chatbots, real-time multiplayer apps and cloud-native platforms on AWS. Currently finishing my B.Sc. in Computer Engineering at the German University in Cairo with an Excellent grade (GPA 1.6, German scale).
attention.py
class SelfAttention(nn.Module):
def forward(self, x):
q, k, v = self.qkv(x).chunk(3, dim=-1)
att = (q @ k.transpose(-2, -1)) * self.scale
return att.softmax(dim=-1) @ v24+Projects shipped
1.6German GPA
3Professional roles
2026Graduating
PyTorch/TensorFlow/Transformers/LangChain/RAG/YOLO/Neo4j/FastAPI/Spring Boot/Flutter/Kubernetes/AWS/Kafka/Docker/React/PostgreSQL/
01Career
Work experience
- Sep 2024 — Dec 2024
- Jul 2024 — Aug 2024
Technical Support Intern
Orange Business
Network security, technical support, cloud partitioning and cloud solution architecture.
- Oct 2023 — Nov 2023
Software Developer Intern
Hotel am Borsigturm, Germany
Built the front-end website with HTML and Tailwind, and developed an automated PDF generator still used for customer documents.
02Toolkit
Technical knowledge
The stack I reach for across research prototypes and production systems.
AI & Machine Learning
13- LLMs
- Transformers
- Neural Networks
- Deep Neural Networks
- Transfer Learning
- Convolutional NNs
- Fine-tuning
- Embeddings
- RAG
- Knowledge Graphs
- Computer Vision
- TensorFlow Lite
- YOLO
Languages
09- Python
- Java
- Dart
- TypeScript
- JavaScript
- C
- SQL
- Lua
- Bash
Frameworks & Libraries
14- PyTorch
- TensorFlow
- LangChain
- Hugging Face
- OpenCV
- Spring Boot
- FastAPI
- Flutter
- Riverpod
- React
- Node.js
- Express.js
- Streamlit
- Tailwind CSS
Data & Storage
10- PostgreSQL
- MongoDB
- Neo4j
- DynamoDB
- Redis
- Elasticsearch
- FAISS
- Firebase
- Supabase
- Power BI
Cloud & Infrastructure
12- AWS (EC2, S3, Lambda)
- CloudFront
- EKS
- CloudFormation
- Kubernetes
- Docker
- Kafka
- RabbitMQ
- gRPC
- Prometheus
- Grafana
- Linux
Engineering Practices
09- Microservices
- Test Driven Development
- RESTful APIs
- OOP
- Design Patterns
- Git & GitHub
- Agile
- SCRUM
- Jira
03Work
Selected projects
Deep learning research, RAG systems, real-time multiplayer products and cloud-native architecture.
A cross-platform mobile party game with 15+ real-time multiplayer modes, built in Flutter with Riverpod on top of a Nakama authoritative realtime server running custom Lua RPC modules for leaderboards, entitlements and chat moderation over WebSocket matches. A Python FastAPI backend on PostgreSQL serves 28,000+ content items alongside a server-authoritative in-app economy, a Gemini-powered AI companion, LiveKit voice rooms and IAP receipt verification for Google Play and the App Store. Also features on-device face tracking with a quantized YOLO model via TensorFlow Lite for a camera-based Quran recitation game, plus a React + TypeScript admin CMS and an event analytics pipeline with partitioned tables and daily rollups. I authored 80% of a 300+ commit codebase in a two-person team.
- Flutter
- Dart
- Riverpod
- Nakama
- Lua
- FastAPI
- +8
Featured
Cloud-Native Web App on AWS
A highly available full-stack CRUD application deployed across multiple EC2 instances in separate Availability Zones behind an Elastic Load Balancer and fronted by CloudFront for low-latency global delivery. DynamoDB stores the dataset while S3 holds versioned image uploads linked to each record, and a Lambda function triggered on item creation resizes images through an SNS/SQS event-driven pipeline. Containerized services run on Amazon EKS, infrastructure is managed with CloudFormation, system health is monitored via CloudWatch, and access is locked down with least-privilege IAM policies and VPC security groups.
- AWS EC2
- Elastic Load Balancer
- CloudFront
- DynamoDB
- S3
- Lambda
- +9
Featured
Freelance Marketplace Platform
A two-sided marketplace connecting freelancers with clients, built as five Spring Boot microservices — User, Job, Proposal, Contract and Payment — each backed by PostgreSQL for transactional data. A polyglot persistence layer adds MongoDB for portfolios and threaded messaging, Neo4j for graph-based skill matching, Elasticsearch for full-text job search with filters, and Redis for caching active listings and sessions. Services communicate over OpenFeign, RabbitMQ, Kafka and gRPC, applying Saga, Observer, Mediator and Builder patterns, and the platform runs on Kubernetes with Prometheus and Grafana monitoring. I served as Scrum Master, driving sprint planning and backlog grooming across the team.
- Spring Boot
- Microservices
- PostgreSQL
- MongoDB
- Neo4j
- Elasticsearch
- +9
Featured
She3rGPT
A generative language model built from scratch with a decoder-only Transformer in PyTorch, trained on classical Arabic poetry to generate new verses in a similar style. The project also includes educational notebooks covering neural networks, evapotranspiration prediction, self-attention, and autoregressive text generation.
- Transformer Architecture
- PyTorch
- LLM
- Deep Neural Networks
- Embedding
- Python
Featured
E-Commerce Graph-RAG Chatbot
Intelligent e-commerce search powered by Knowledge Graphs and Large Language Models, combining sentiment analysis with graph-based retrieval.
- Neo4j Knowledge Graph
- LangChain
- HuggingFace LLMs
- Streamlit
- Python
A machine-learning web application predicting evapotranspiration from 15 years of historical and forecast weather data, with a RAG layer giving irrigation recommendations tailored to location, crop and soil type — supporting sustainable agriculture through data-driven decisions.
- Deep Neural Networks
- TensorFlow
- RAG
- API
- Vercel
- Render
Showing 6 of 24
04Education
Academic background
Aug 2021 – 2026 (Expected Graduation)
German University in Cairo
B.Sc. Computer Engineering — Faculty of Media Engineering and Technology
German GPA
1.6
Excellent
US GPA (4.0 scale)
3.7
A−
Full résumé
Youssef Magdy
05Teaching
Professional highlight
Loading playlist details…
YouTube playlist about Transformers
I built a YouTube learning playlist focused on Transformers and practical AI concepts — breaking down attention, embeddings and model internals for people getting started.
Watch on YouTube06Contact
Let's build something
I am always happy to talk about AI engineering, machine learning research or an interesting product idea.