📧 Email • 💼 LinkedIn • 🔗 GitHub • 🐳 Docker Hub
LLM Engineer, AI Backend Engineer, RAG Systems, AI Agents, LangGraph, Pydantic-AI,
Multi-Agent Systems, Prompt Engineering, Function Calling, FastAPI, Python Backend,
Distributed AI Systems, Microservices Architecture, Kubernetes, Docker, PostgreSQL,
pgvector, Vector Databases, Qdrant, Celery, RabbitMQ, Redis, Asynchronous Programming,
MLOps, Model Deployment, Fine-tuning (QLoRA), PyTorch, TensorFlow, NLP, Transformers,
Tool Orchestration, Autonomous Reasoning, Retrieval-Augmented Generation, Embeddings,
CI/CD, Prometheus, Grafana, Cloud Deployment, Embedded Linux, AUTOSAR
Backend Engineer specializing in AI-powered systems and agentic architectures.
I build production-grade, scalable AI backends with a focus on:
✅ Agentic AI Systems — Multi-agent orchestration with LangGraph, tool usage, and autonomous reasoning
✅ RAG Pipelines — Multi-step retrieval, query decomposition, and embeddings-based search
✅ Scalable FastAPI Services — Async processing, distributed task queues, and low-latency inference
✅ Production Infrastructure — PostgreSQL, vector databases (pgvector, Qdrant), message brokers (RabbitMQ)
✅ Observability — Prometheus, Grafana, monitoring, and alerting for distributed AI systems
Currently expanding into: Agentic AI systems with LangGraph, tool orchestration, multi-step reasoning, and autonomous, production-ready AI services.
Unique Background:
- AI + Backend Engineering fusion
- Embedded Systems experience → production reliability mindset
- Industrial engineering background → process optimization and quality focus
- AI Agents: LangGraph, Pydantic-AI, tool orchestration, multi-agent systems
- LLM Engineering: Prompt orchestration, function calling, autonomous reasoning
- Agentic Workflows: Multi-step reasoning, critique-based refinement, structured state management
- Prompt Engineering: Few-shot, chain-of-thought, agentic prompting, evaluation
- Retrieval-Augmented Generation (RAG)
- Multi-step retrieval & query decomposition
- Embeddings: OpenAI, sentence-transformers, similarity search
- Vector Databases: Qdrant, pgvector
- LLM Evaluation & Latency Optimization
- FastAPI (async, high-performance APIs)
- RESTful API Design
- Asynchronous Programming
- Microservices Architecture
- Message Brokers: RabbitMQ
- Task Queues: Celery
- Pydantic Data Validation
- SQLAlchemy ORM
- Database Design
- Docker & Kubernetes (container orchestration)
- Cloud Deployment (AWS, GCP)
- CI/CD Automation (GitHub Actions)
- Observability: Prometheus, Grafana
- Monitoring & Alerting
- Redis (caching, pub-sub)
- Model Deployment & Serving
- PostgreSQL (production-grade relational DB)
- pgvector (vector search in Postgres)
- MongoDB (NoSQL)
- Qdrant (vector database)
- Indexing & Query Optimization
- PyTorch, TensorFlow
- Fine-tuning (QLoRA)
- Transfer Learning
- NLP & Transformers
- Computer Vision (OpenCV)
- Hyperparameter Optimization
- Embedded Linux (Yocto, Buildroot)
- AUTOSAR
- ARM Architecture, RTOS
- QEMU Emulation
- Cross-Compilation, CMake, Make
B.Sc. in Mechanical Engineering | Banha University | 2016 – 2022
🏆 Ranked 5th out of 120 in the Mechanical Engineering Department
- AI Engineer Core Track: LLM Engineering, RAG, QLoRA, AI Agents | Udemy | 2026 (in progress)
- Prompt Engineering and RAG | Eng. Abu Bakr Soliman | 2026
- Computer Vision with Python, TensorFlow, and PyTorch | freeCodeCamp | 2025
- CNN, RNN, LSTM, GRU, NLP, and Transformers | Stanford University | 2024
- Statistics, Linear Algebra, and Numerical Optimization | Dr. Hatem Elattar | 2024
- Machine Learning and Deep Learning Diploma | Eng. Hesham Asem | 2024
- Fundamentals of Backend Engineering | Udemy | 2025
- FastAPI Development, Data Structures, Algorithms, and OOP | freeCodeCamp | 2024
- Database, CCNA, Cloud Computing, Kubernetes, Jenkins, and Redis | Udemy | 2025
- AUTOSAR Diploma | IMT School | 2024
- Embedded Linux, CMake, Google Test, GDB, QEMU | Eng. Moatasem Elsayed | 2024
- Embedded System Diploma | Eng. Ahmed Abd El-Ghafar | 2023
- Lean Six Sigma Yellow and Green Belt | Udemy | 2022
02/2025 – 02/2026
- Led cross-functional training and process optimization initiatives for washing machine assembly lines
12/2023 – 12/2024
- Applied RCA, DMAIC, and DFMEA methodologies to diagnose and resolve failures on VW and Renault assembly lines
03/2023 – 12/2023
- Responsible for continuous improvement activities for Water Dispenser and TV production lines
Advanced multi-agent LLM pipeline demonstrating production-grade agentic architecture:
Key Features:
- 5-Agent Pipeline: Planner → Researcher → Writer → Critic → Finalizer
- Iterative Reasoning: Structured state management with conditional feedback loops
- Self-Improving Workflow: Critique-based refinement for higher response quality
- Structured Outputs: Pydantic models for type-safe agent communication
- Production-Ready: Error handling, logging, and observable state transitions
Tech Stack: LangGraph, Pydantic, OpenAI API, Structured Outputs
Impact: Achieved measurably higher response quality vs. single-agent baselines through multi-stage refinement
Production-grade distributed AI backend for document intelligence:
Architecture:
- Distributed Backend: FastAPI + Celery + RabbitMQ for async workloads
- Retrieval + Reasoning Pipeline: LLMs with vector search (Qdrant/pgvector)
- Async Task Queues: Celery workers for scalable, decoupled services
- Agentic Workflows: Tool usage and multi-stage execution for complex reasoning
Performance Optimizations:
- ⚡ <150ms API latency (8.6x improvement over synchronous blocking)
- Async endpoints, non-blocking I/O, Redis caching, connection pooling
Observability:
- Prometheus + Grafana dashboards
- Real-time monitoring: task queues, vector DB performance, LLM latency
Tech Stack: FastAPI, Celery, RabbitMQ, PostgreSQL, pgvector, Qdrant, Redis, Prometheus, Grafana
Custom Embedded Linux system for edge AI inference:
Key Achievements:
- Built custom Embedded Linux distribution using Yocto Project for Raspberry Pi 4
- Optimized kernel and userspace for AI inference workloads
- Deployed containerized Whisper AI model for real-time audio transcription
- Utilized QEMU emulation for rapid dev iteration before hardware deployment
- Integrated OpenCV for computer vision capabilities
Tech Stack: Yocto, Buildroot, Whisper, Docker, QEMU, OpenCV, Raspberry Pi 4
Impact: Demonstrated model deployment expertise and edge computing optimization strategies
Production-ready transfer learning pipelines:
Implementation:
- Pre-trained models: ResNet, VGG, EfficientNet
- Fine-tuning strategies with data augmentation
- Hyperparameter optimization for production accuracy
- Model evaluation and performance benchmarking
Tech Stack: PyTorch, TensorFlow, torchvision, scikit-learn
NLP-based bilingual spam detection system:
Features:
- Bilingual support: Arabic and English
- Advanced preprocessing (stemming, lemmatization, stopword removal)
- Multiple vectorization methods: Count Vectorizer, N-grams, TF-IDF
- Custom feature engineering
- Random Forest classifier with cross-validation
- Feature importance analysis
Tech Stack: NLTK, scikit-learn, pandas, NumPy
Data-driven customer segmentation for automotive dataset:
Techniques:
- KMeans clustering for customer segmentation
- Random Forest and Linear Regression for predictive modeling
- Feature engineering and outlier detection (DBSCAN)
- Dimensionality reduction (PCA)
- Actionable business insights extraction
Tech Stack: scikit-learn, pandas, NumPy, matplotlib, seaborn
Full embedded software stack for AVR microcontrollers:
Architecture:
- Layered Design: MCAL → HAL → Service → Application
- Reusable drivers and structured layers
- Support for AVR builds and Proteus simulation
Test Applications:
- Keypad-to-LCD interface
- Interrupt-driven LED control (EXTI)
- I2C master-slave communication
- Multi-ultrasonic sensor distance measurement with LCD display
Additional Features:
- Custom linker script and startup code for ARM target
- CMake/Make build automation
- Doxygen technical documentation
Tech Stack: C, AVR, ARM, CMake, Make, Doxygen, Proteus
🔗 GitHub Repository (Graduation Project)
Advanced robotics simulation and control system:
Implementation:
- Modeled 6-DOF robotic manipulator in MATLAB/Simulink
- Forward and inverse kinematics for precise end-effector positioning
- Path planning and trajectory generation algorithms
- Control laws for smooth, stable, and accurate motion control
Tech Stack: MATLAB, Simulink, Robotics Toolbox
Languages: Python, C, C++, SQL, Bash
AI/ML Frameworks: PyTorch, TensorFlow, LangChain, LangGraph, Pydantic-AI
Backend: FastAPI, SQLAlchemy, Pydantic, Async Python
Databases: PostgreSQL, pgvector, MongoDB, Qdrant, Redis
Infrastructure: Docker, Kubernetes, RabbitMQ, Celery
MLOps: CI/CD (GitHub Actions), Prometheus, Grafana
Cloud: AWS (RDS, S3, EC2), GCP
Embedded: Yocto, Buildroot, AUTOSAR, RTOS, ARM, QEMU
Tools: Git, Postman, STM32CubeMX, Cursor, Jira, Trello, Power BI
- Agentic AI Systems: Multi-agent orchestration, tool usage, autonomous reasoning
- Production RAG: Scalable retrieval pipelines with distributed backends
- Kubernetes-Based AI Deployment: Container orchestration for AI workloads
- Observability: Monitoring and optimizing distributed AI systems
- Arabic: Native
- English: Professional Working Proficiency
- German: Basic
I'm always open to discussing AI engineering, backend architecture, and distributed systems.
- 📧 Email: ahmed.ferganey707@gmail.com
- 💼 LinkedIn: linkedin.com/in/ahmed-ferganey
- 🔗 GitHub: github.com/ahmedferganey
- 🐳 Docker Hub: docker.com/u/ahmedferganey
- 📍 Location: Al-Qalyubia, Egypt
- 📱 Phone: +20 1091657144
Building the future of AI-powered systems, one microservice at a time.