Health Tech
B2B
Clinical Decision Support
Applied AI
*The underlying ML model was research-based and not deployed in production
*Redesign completed independently to explore how such a system could function as a real clinical decision-support tool.
In semi-urban clinics, doctors and lab technicians work under time pressure, limited infrastructure, and high cognitive load. While machine-learning models can help detect liver disease early, they only add value if clinicians can understand, trust, and act on the results.
This project began with an accurate ML prototype that technically worked but failed to fit real clinical workflows. My role was to redesign the experience so the model functioned as a practical decision-support tool, not just a prediction engine.
Explainable Prediction
Making risk defensible
Binary predictions highlighted abnormal values but gave no context clinicians could trust or explain.
The prediction view intentionally surfaces only abnormal values to reduce noise, while clearly communicating risk level and contributing factors so clinicians can quickly understand what matters and why.
In-Context Chatbot
Resolving uncertainty on demand
Clinicians often needed to understand normal ranges, value significance, or implications that weren’t visible on the main screen.
Static explanations would have cluttered the interface and slowed workflows.
The chatbot acts as an on-demand interpretation layer allowing clinicians to ask about any value, definition, or threshold without leaving the flow or overloading the UI.
Embedded Appointment Booking
Turning insight into access
In remote and semi-urban clinics, identifying risk isn’t enough access to the right specialist is often the real bottleneck.
Appointment booking is embedded directly into the decision flow, helping clinicians connect patients with appropriate specialists immediately, rather than relying on fragmented or offline referrals.
Outcome
