AI Engineer specializing in LLMs, Retrieval-Augmented Generation (RAG), and NLP, with experience building production-ready intelligent document processing systems. Focused on developing enterprise-grade AI systems that d…
AI Engineer specializing in LLMs, Retrieval-Augmented Generation (RAG), and NLP, with experience building production-ready intelligent document processing systems. Focused on developing enterprise-grade AI systems that deliver measurable business impact. Skilled in Python, PyTorch, LangChain, FAISS, and deep learning, with experience developing scalable AI solutions including LLM + OCR pipelines, semantic retrieval systems, and machine learning models for real-world automation and analytics. Passionate about building practical AI applications that deliver measurable impact.
Developed advanced recommendation systems using quaternion embeddings to improve sparse data representation and similarity learning.
Architected and implemented a custom Quaternion Matrix Factorization model in PyTorch to enhance representation of sparse user–item interactions in recommendation systems.
Developed and optimized Hamilton product-based embedding interactions using dual projection methods (radius and angle) to improve similarity learning.
Optimized embedding dimensions (K=4–32), achieving peak performance with RMSE of 0.537 and MAE of 0.455.
Engineered an end-to-end recommendation pipeline, including model training and Top-N recommendation generation.
Leveraged PyTorch, NumPy, and efficient batching techniques to enable scalable large-scale prediction.
Developed an AI-powered document validation system using LLMs, OCR, and Retrieval-Augmented Generation (RAG) techniques to automate extraction and verification of structured banking data. Built an agentic RAG pipeline with semantic retrieval and embeddings, achieving 83.3% field-level extraction accuracy with an F1-score of 0.91 while reducing manual review effort by 40%. Deployed the system using Streamlit for real-time document processing.
Developed an advanced recommendation system using quaternion embeddings to improve sparse data representation and similarity learning. Implemented a custom Quaternion Matrix Factorization model in PyTorch, optimized embedding interactions using Hamilton products, and built an end-to-end recommendation pipeline including model training, prediction, and Top-N recommendation generation. Achieved strong performance with RMSE of 0.537 and MAE of 0.455 using scalable batching and efficient prediction techniques.
نظام التوظيف التشغيلي المعتمد على الذكاء الاصطناعي أولاً، الذي يوحّد ويؤتمت دورة التوظيف بالكامل للمؤسسات الحديثة حول العالم.
د.جوب منطقة حرة ذ.م.م. 2026 © جميع الحقوق محفوظة
