Role
Product Designer · Frontend Developer
Earlier clinical ML work: a dashboard concept for transplant physicians that made scattered patient data, temporal changes, and prediction confidence easier to interpret.
Artifact pass pending before public launch: replace placeholders with dashboard screenshots, research synthesis, and usability-result visuals.
Clinicians can identify acute rejection or infection quickly, but detecting long-term complications is far harder. Physicians depend on continuous review of medical data to spot patterns that might signal life-threatening outcomes years after transplantation.
EMR systems store patient data in broad profiles, yet specialized departments struggle with information overload. Much of the collected data is irrelevant to their context — consuming clinician and patient time while obscuring the medical changes that actually matter for transplant monitoring.
A specialized dashboard presenting targeted patient data with temporal changes reduces navigation friction for medical teams. Machine-learning integration enables prediction of long-term complications, strengthening post-transplant surveillance.
“Data is often scattered across our system and searching for it wastes my time.”
— Clinician interviewI began with expert interviews and scholarly literature to grasp the complexities of organ-transplant recovery. This established foundational knowledge about how medical history affects post-surgical outcomes — and revealed clinician frustration with fragmented data collection and analysis.
Through observation of clinicians and nursing staff, the team documented daily workflows and system interactions, producing practical insight into unmet needs. Combined with prior interviews, this yielded an affinity map synthesizing developer and end-user concerns.
The team examined multiple EMR systems deployed at UHN, building understanding of the existing clinical infrastructure and integration pathways.
Field observations revealed a concerning pattern: six of ten male patients showed non-compliance with prescribed treatments.
This informed a patient-centered visualization encouraging treatment adherence — approved by both stakeholders and clinicians.
Feedback revealed divergent preferences: senior clinicians preferred date-organized views (reflecting familiarity with legacy EMRs), while junior practitioners favored parameter organization. This anchoring bias prompted a customizable table orientation.
Summary review and editing capabilities were integrated, allowing corrections before finalization — making temporal trends legible at a glance.
Communicating prediction accuracy required intuitive Feature Importance presentation. Since most clinicians lack ML literacy, the team used A/B/C testing and Think-Aloud protocols to refine clarity.
Ten previously uninvolved clinicians completed comparative testing — identifying a specific patient parameter value on date B, then comparing it against the same parameter on date C.
Results showed the prototype substantially outperformed the legacy EMR system for task-completion efficiency.
The project strengthened three product-design competencies:
Clinician collaboration was the greatest challenge. Rapid-prototyping tools (Balsamiq, FigJam, Figma) made iteration fast, while Miro enabled organized feedback synthesis across design, development, and clinical teams.
I'm here to listen, collaborate, and design solutions that make impact.
Get in touch! →