Quantum Bit

Built and deployed an end-to-end system that enables users to interact with their own data via API. The system accepts raw data through an API endpoint (/data), automatically cleans, extracts, and stores it in a structured database. Leveraging Retrieval-Augmented Generation (RAG) and a fine-tuned language model, it allows users to query and chat with their data through a second API endpoint (/chat). This solution combines data processing, vector search, and language modeling to deliver intelligent, context-aware responses in real-time.

Tools: LLMs - RAG – LangChain – LlamaIndex – Vector Databases – Chroma – Prompt Engineering – Text Embedding – Data Cleaning – NLP

During the Hajj season in Saudi Arabia, many pilgrims from different nationalities arrive to perform Hajj, and a large number of emergency cases occur. In many situations, callers cannot be understood because they speak different languages or speak very fast. Because of this problem, the Saudi Ministry of Hajj and Umrah contacted me to build a system that translates speech in real time on both sides of the call.

I built a fast and powerful system using open-source AI models that supports Arabic, English, French, Urdu, and Indonesian, and I also added a call analysis module that stores each emergency call in the database after it ends.

Tools: OpenSource -Finetuning - (TTS, LLMs, STT) - Websocit - Llama.cpp

Built a Text-to-SQL AI Agent that converts natural language prompts into accurate SQL queries using a dynamic database schema. Developed custom tools to securely execute queries, retrieve data, and visualize results in a user-friendly way.

Implemented authentication, role-based access control, and data, level permissions, ensuring users can only access their authorized data, enabling safe and intuitive database interaction without SQL knowledge.

Tools: FastAPI - Chart.js - Custom tools - LangGraph - Chroma Server

Developed a vision system for a self-driving car tailored to Egyptian streets using over 5,000+ locally collected and online images. Designed, built, and tested 6 specialized AI models combining object detection, generative AI, and reinforcement learning to handle lane following, obstacle avoidance, and traffic sign recognition. Achieved 92%+ object detection accuracy with reliable real-time decision-making during on-street testing across multiple traffic scenarios.

Tools: OpenCV – TensorFlow – PyTorch – Computer Vision – Generative AI

Built a multi-agent system where autonomous agents handle web search, summarization, and email delivery from a single user prompt. Used Serper for real-time search, CrewAI and AutoGen for orchestration, and deployed the interface with Streamlit.

Tools: CrewAI – AutoGen – LangChain – Serper – SMTP – Streamlit

Bachelor of Computer Science

Al Specialization

Web API - MVC - DataBase - C# - Angular