Almas Hilman Muhtadi
This project sits at the intersection of software engineering, state-of-the-art AI agent
technology, and healthcare, addressing key industrial and clinical challenges in hospital report
automation. Through close collaboration among clinical, product, and technical stakeholders,
we are developing ARCi (Agentic Rule Creation System). This AI agent framework empowers
clinical application specialists to generate clinical rules directly from clinical specifications,
turning a complex rule-creation process into a streamlined process.
The main challenges I solved in this project were that general AI models aren't trained on
building clinical rules using domain-specific language, and that each hospital uses its own
terminology and guidelines. So the main challenge was steering AI agents to write technically
correct, consistent clinical rules in a precise format amid changing context. The workflow
therefore had to let the agents check and fix their own output. We also had to monitor output
quality each time we tuned the agents.
The project delivered ARCi as an early-stage proof of concept. Agent skills hold the technical
expertise, and the agents base each rule on a hospital-specific clinical guideline and a tool that
resolves each hospital's terminology. Through a UI, the application specialists keep control of
the result. To monitor quality, an evaluation pipeline scores agent performance and how closely
its generated code matches a human reference rule. This score tracks agent performance over
time.
The final result is ARCi, which can produce clinical rules at least 6 times faster and 30 times
cheaper than the manual process. This removes rule-customization bottlenecks and enables
scalable automated reporting across hospitals. This lets clinicians produce higher-quality
reports faster, directly improving overall patient care.

