A Senior Data Scientist in Applied Research developing GenAI NLP products. Managed BI & Analytics, overseeing dashboarding and reporting, and assisting with data governance. I spent most of my career as a web developer, building enterprise applications, during which I completed my Master's in Artificial Intelligence. My foundation began with a Bachelor in Software Engineering, and throughout my career, I've been passionate about bridging technical teams with business users, mentoring developers, and delivering high-impact solutions.
Working in Applied Research, building intelligent products including smart contract assistance, document summarization, legal research and analysis, and conversational interfaces. Developing full-stack AI applications using Python, FastAPI, LangChain, LangGraph, PostgreSQL, Milvus, and Redis with RAG pipelines. Leading vibe-coding strategy across the team—establishing best practices for handling generated artifacts, PR reviews, skill development, and aligning tooling for a unified development experience.
Led end-to-end BI solution development and governance across multiple domains using Power BI, SAP BO, and Looker; streamlined demand management and workflow processes via Azure DevOps and IT ticketing systems; and built a Dialogflow CX–Vertex AI conversational agent specializing in Saudi data governance and banking regulations.
Coursework spanning Foundations of AI, Machine Learning, Deep Learning, Mathematics for Data Science, Computer Vision, Natural Language Processing, and Evolutionary Computation & Global Optimization. Developed a solid understanding of both theoretical and applied AI, with an award-winning project in Compliance Matching for Retail Shelf Monitoring, recognized as the university's Best Project. The work involved building an intelligent shelf monitoring system for out-of-stock detection using advanced deep learning and computer vision techniques, published in ICONIP2024, and another paper published at ArabicNLP 2024.
Served as Lead Developer, Technical Advisor, Business Analyst, and Application Support specialist—leading workflow system development and rock sample management enhancements, delivering CI/CD-focused training, supporting RemoteQC deployment and database operations, and ensuring efficient, user-centered application performance across divisions.
Developed an approval workflow platform for publications, patents, and recognition. Led backend and frontend development, SAP workflow integration, legacy data migration, and automated testing. Delivered high-availability support up to executive level.
Modernized two legacy modules and developed multiple components for managing the operations of one of the world's largest core storage facilities. Technologies included Java, Angular, and Oracle with a focus on scalability and maintainability.
Built front-end dashboards for a seismic and non-seismic workflow management platform. Focused on enhancing usability, readability, and operational reporting.
Development of a seismic field reporting platform analysing equipment status data at scale. Created simulations for performance tuning and delivered Angular training across the division.
Provided frontline support for seismic interpretation tools (Petrel, GeoFrame, etc.). Also managed deployments, licensing, and developed internal training content.
The program provided a solid foundation in software development and engineering principles. Covered the full software lifecycle, including requirements analysis, design, implementation, testing, and maintenance. Combined strong theoretical grounding in mathematics and computer science with practical experience through team projects and industrial training, emphasizing quality, collaboration, and modern development practices.
Thanks to everyone who I worked with to achieve these.
July 2025
The paper introduces RetailEye, a two-stage deep learning system using YOLOv8s and EfficientNetV2-S/ResNet18 that leverages supervised contrastive learning to accurately detect and recognize grocery products in complex retail environments, outperforming one-stage methods and providing a new custom dataset for inventory management research.
August 2024
The study evaluates three large language models on the Arabic stance detection dataset MAWQIF and finds that fine-tuning, particularly with GPT-3.5-Turbo achieving a macro-F1 score of 82.93, significantly improves performance over prior BERT-based models, highlighting the promise of LLMs for Arabic language tasks.
March 2018
The paper presents an automated quality control (QC) system for seismic acquisition operations that integrates diverse data types into a centralized database, using parallel processing, data normalization, and on-the-fly error correction to ensure consistent, efficient, and scalable monitoring of equipment performance and operational compliance.