Digital Engineering • AI Safety
This project encompasses the core programming problem-solving assignments for the MSc Software Engineering modules at City St George's, University of London. The implementation pipeline utilizes Python for backend logic, OOP (Object-Oriented Programming) for system architecture, and is integrated within a full-stack environment using HTML, CSS, and JavaScript to ensure robust and scalable solutions.
View Technical Solutions
Redesigned the bone densitometry medical report stylesheets within Nuclear Medicine GE reporting to deliver specialized, structured reports aligned with WHO classification standards. In addition, engineered a critical architectural workaround in pediatric report core logic to dynamically calculate pediatric skeletal age—bypassing native system limitations to compute precise, custom Z-scores and securing 100% data accuracy across Saudi Aramco clinical networks.
Architected the regional deployment of Clinical Decision Support (CDS) for Radiant-Nuclear Medicine. Innovated system architecture by redesigning traditional single-visit workflows into multi-visits integrated procedures pipelines. Led the migration of 200+ mission-critical workflows, ensuring flawless interoperability, state preservation, and data integrity between PACS, RIS, and SAP EHR ecosystems.
Developing robust benchmarking protocols for Large Language Models (LLMs) within clinical environments by executing rigorous comparative analyses of output behaviors across competing AI models on complex healthcare case studies. Operating as a Medical Expert and AI Trainer at Outlier, this research establishes strict safety standards, optimization strategies via structured medical prompting, and auditing metrics under adversarial pressure to ensure compliance with global healthcare guidelines.
Lead Researcher for the Clinical Prompt Injection Protocols (CPIP) validation phase. This research is specifically engineered to mitigate clinical hallucinations where AI prematurely "guesses" answers without requesting missing medical data.
By shifting the interaction from generic chat to a specialized reasoning simulation, CPIP mandates iterative differential refinement. It forces the AI to preserve diagnostic uncertainty and actively query for critical patient parameters, bridging the gap between conversational models and safety-critical medical decision-support.
Demonstrating execution of design thinking, visual hierarchy, and UI/UX prototyping paradigms translated into real-world educational modeling. This section showcases advanced leadership and instructional architecture by mentoring early-stage creators to build high-fidelity interactive models and prize-winning nutritional visual assets.