Building production-ready AI systems powered by Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Multi-Agent AI, and Machine Learning.
I'm an AI Engineer passionate about building intelligent systems that bridge cutting-edge AI research with real-world applications.
My expertise spans Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Multi-Agent Systems, LLM Fine-Tuning, Machine Learning, and Production AI Engineering.
I enjoy designing scalable AI architectures, developing intelligent automation workflows, and deploying production-ready AI applications using modern engineering practices.
- AI & LLMs: Large Language Models • Agentic AI & Multi-Agent Systems • Retrieval-Augmented Generation (RAG) • LLM Fine-Tuning (QLoRA / LoRA)
- Engineering: FastAPI & Backend Engineering • Docker & Production Deployment
- Data: Machine Learning & Data Science
🎓 Soual — AI-Powered Question Generation Platform (Graduation Project)
An AI platform that generates Egyptian Ministry-style examination questions using Large Language Models. My part: the AI question-generation system.
- Curriculum-aware Retrieval-Augmented Generation (RAG)
- LLM fine-tuning (LoRA) for educational question generation
- Semantic search with vector databases
- Automated PDF parsing and knowledge extraction
- Structured JSON output: multiple-choice, true/false, and short-answer questions with distractors and explanations
- Prompt engineering and output validation
- FastAPI REST API and Docker deployment
🔗 Code: github.com/AI-Question-Generator/question-generation-system
🤗 Models: question_generation_1.5B_model_v2 • question-generator-model-qwen2.5-1.5b • all models
Tech Stack: FastAPI • LangChain • Transformers • Hugging Face • PEFT / LoRA • PostgreSQL (pgvector) • Docker
🧠 RAGGuide_ — Production-Ready RAG Application
A retrieval system with two modes: Q&A for grounded answers, and Judge Mode, which checks documents against rules or policies and returns structured improvement feedback.
- Modular FastAPI backend with a provider factory to swap OpenAI or Cohere
- Semantic chunking, embeddings, and vector similarity search
- PostgreSQL + pgvector, with Qdrant supported
- Docker Compose deployment
- Prometheus metrics and Grafana dashboards
🔗 Code: github.com/ahmdeltoky03/RAGGuide_
Tech Stack: FastAPI • PostgreSQL • pgvector • Streamlit • Docker • Prometheus • Grafana
🏢 Company Report Generator — Multi-Agent Research System
Give it a company name and it researches the web, structures the findings, and writes a professional business report.
Research Agent → Analysis Agent → Writer Agent → Company Report
- CrewAI orchestration with parallel Tavily searches and cited sources
- Pydantic-validated JSON from every agent
- Report streamed live to a lightweight web UI
- Markdown and PDF export
🔗 Code: github.com/ahmdeltoky03/company-report-generator • 🎬 Demo video
Tech Stack: CrewAI • FastAPI • Cohere • Tavily • Pydantic
⚖️ Avokat AI — Legal Assistant with GraphRAG (NTI Summer Internship)
Upload a legal PDF and it builds a knowledge graph of the entities and relationships inside, then answers questions grounded in that document, in Arabic, English, or a mix.
- Neo4j knowledge graph built with LangChain
- PyMuPDF ingestion with automatic language detection
- Token-by-token streaming chat (Server-Sent Events) on Gemini 2.5 Flash
- Isolated graph per chat session, with source citations
- React interface over a FastAPI backend
🔗 Code: github.com/ahmdeltoky03/Avokat-AI • 🎬 Demo video
Tech Stack: FastAPI • React • Neo4j • GraphRAG • Gemini • Sentence Transformers
⚽ Sports Analytics Platform — Multi-Agent Football AI
Ask a football question in Arabic or English and six specialized agents fetch live data, analyze it, and return a polished report.
- CrewAI orchestration with shared context memory
- Football-Data API integration
- Gemini and Groq as interchangeable model backends
- Exports to Markdown, JSON, and PDF
🔗 Code: github.com/ahmdeltoky03/agentic-system-crewai
Tech Stack: CrewAI • Gemini • Groq • FastAPI • Football-Data API
🛒 Smart Procurement Agent — Automated Product Research
Four agents turn a shopping request into a procurement report: they craft search queries, search the web, scrape product pages, and compare prices and specs.
- Search-query, search-engine, scraping, and report-author agents in sequence
- HTML report generated automatically
🔗 Code: github.com/ahmdeltoky03/smart-procurement-agent • 📄 Live example report
Tech Stack: CrewAI • OpenRouter • Tavily • ScrapeGraph
| Project | What it does | Stack |
|---|---|---|
| Retail Analytics Copilot | Fully local RAG + text-to-SQL agent; DSPy tuning lifted valid-SQL rate from 60% to 85% | LangGraph DSPy Ollama |
| AI Medical Chatbot | Medical Q&A bots fine-tuned on BioMistral-7B and 4-bit LLaMA 3.2 | Unsloth Streamlit |
AI & LLM: RAG • GraphRAG • Multi-Agent Systems • CrewAI • LangGraph • LangChain • DSPy • Prompt Engineering • LoRA / QLoRA • Evaluation
Backend: Python • FastAPI • REST • Pydantic
Data stores: PostgreSQL • pgvector • Qdrant • Neo4j
Deploy & Monitor: Docker • Docker Compose • Prometheus • Grafana
Turning ideas into intelligent systems through AI engineering, scalable architectures, and production-ready solutions.