Building AI Systems that transform ideas into real world solutions.
AI Engineer specialized in Machine Learning, Deep Learning, and Generative AI systems.
About Me
Hi, I'm Omar Khalil, an AI Engineer and Informatics Engineering graduate from Arab International University (AIU). My journey started with machine learning and deep learning, and has expanded to include computer vision, generative AI, Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and AI Agents.
I enjoy building end-to-end AI projects, from data processing and model development to deploying them as scalable, production-ready applications. Beyond my technical work, I share my knowledge by creating specialized AI content on LinkedIn, aiming to simplify modern concepts and share the latest technologies and practical projects with the tech community.
I am currently open to job opportunities, internships, and collaborations in AI, computer vision, and generative AI, where I can learn, contribute, and work within teams building technical solutions that make a real impact.
Experience
Freelance AI Engineer
- Engineered end-to-end AI systems spanning Computer Vision, RAG applications, and multi-agent workflows using PyTorch, FastAPI, LangChain, and LangGraph.
- Developed and deployed production-ready microservices for banking systems, document intelligence, and multilingual receipt recognition with high-throughput processing.
- Integrated object detection, OCR engines, vector databases (ChromaDB), and local/cloud LLM inference (Ollama, Gemini API) into scalable pipelines.
All Skills
Machine Learning
- Python
- Pandas
- Data Visualization
- Data Preprocessing & Cleaning
- Feature Engineering
- Scikit-learn
- Hyperparameter Tuning
- Model Selection
- Model Evaluation
Deep Learning
- PyTorch
- Neural Networks
- Computer Vision
- Object Detection
- Ultralytics(YOLOv8)
- Transfer Learning
- Fine Tuning
Generative AI & Advanced Agents
- Large Language Models (LLMs)
- Transformers
- Prompt Engineering
- Retrieval-Augmented Generation (RAG)
- Vector Databases & Embeddings
- AI Agents (LangChain & LangGraph)
Backend Development
- Node.js
- Express.js
- MongoDB
- Mongoose
- SQL
- Database Design
Tools & Frameworks
- Jupyter Notebook
- Google Colab
- Git & GitHub
- Docker
- REST APIs
- FastAPI
- Postman
Soft Skills
- Problem Solving
- Communication
- Continuous Learning
- Teamwork & Collaboration
- Time Management & Organization
- Innovation & Creativity
Featured Projects
A snapshot of my most impactful work in machine learning, computer vision, generative AI, and end-to-end AI systems.
Venezia Bank – AI-Powered Digital Banking System
Venezia Bank is a comprehensive AI-powered digital banking ecosystem developed as a senior graduation project. The platform provides a complete digital banking environment for both customers and bank employees, combining customer-facing mobile and web applications with staff and administrative portals for managing banking operations.
My Contributions
- •Developed an AI-powered financial chatbot using LLMs and RAG.
- •Developed a receipt recognition pipeline for extracting structured financial data from receipt images.
- •Developed a customer churn prediction and retention recommendation system.
- •Developed a financial savings prediction and recommendation system.
- •Integrated AI solutions into the banking platform using FastAPI and REST APIs.
- •Contributed to the initial system architecture, database design, data modeling, and definition of core workflows during the planning and design phase.
Research Contributions
- •Co-authored a research paper presented at the 8th IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies (AEECT 2026), based on a dual-model framework developed as part of this system for stock price movement prediction using technical indicators and financial news sentiment with decision-level fusion.
AI Solutions
8+
User Platforms
3
AI Services
FastAPI
Architecture
Multi-Service
Receipt Recognition System
An end-to-end production system that transforms receipt images into structured financial data. The pipeline combines image validation, region detection, text extraction, and product classification into a unified FastAPI service capable of processing receipts at scale.
Image Validation Accuracy
98.83%
Region Detection mAP50
0.7693
Product Classification F1
0.93
DocMind AI (Multi-Agent)
An intelligent document analysis platform that lets you engage in an expert dialogue with your files. DocMind AI leverages a multi-agent framework to process complex documents, reason through information, and provide accurate, context-aware answers to your queries with high precision.
Architecture
Multi-Agent
Inference
Local/Offline
Retrieval
Hybrid (Vector+Keyword)
Banking Finance Chatbot
An intelligent chatbot tailored for the banking sector based on RAG technology. It integrates vector storage via ChromaDB with advanced large language models (LLMs) to provide accurate and domain-specific financial responses to customer inquiries while maintaining professional banking standards, and features a local fallback path to ensure uninterrupted service.
Vector DB
ChromaDB
Embedding Model
all-MiniLM-L6-v2
BERTScore
88.29%
Customer Churn & Recommendation System
An intelligent banking retention system deployed as a real-time REST API. The system analyzes customer behavior to predict churn risks; if a high risk is detected, it automatically triggers a smart recommendation engine that filters out currently owned products to suggest new, personalized financial solutions, while generating customized marketing messages to ensure customer loyalty.
Models
Dual Pipeline
First Pipeline F1
0.88
Second Pipeline F1
0.88
Menu Classification System
A complete end-to-end machine learning application classifying restaurant images into 5 categories. Demonstrates the entire ML lifecycle: model training with transfer learning, REST API deployment with FastAPI, and interactive web UI with Streamlit for real-time predictions.
Categories
5
F1-Score
0.92
Test Loss
0.2281
NutriVision: Food Classification
A deep learning project comparing various computer vision approaches for food image classification. The system evaluates the performance of state of the art Transfer Learning models (ResNet50 & EfficientNet-B0) against a custom-built AdvancedFoodCNN. It demonstrates the impact of input resolution, data augmentation, and architectural choices on model generalization and training efficiency.
Categories
30
Best Accuracy
79.18%
Top Model
ResNet50
Projects Gallery
More projects, more depth
A broader look at the AI work I've built across research, deployment, and experimentation.
Certifications
Certificate of Authorship – IEEE AEECT 2026
1 / 11
Contact Me
I'm always interested in discussing meaningful AI projects, scalable AI solutions, collaborations, or new ideas for real-world impact.