Context
Telefonica R&D, in collaboration with Hospital Clinic Barcelona. Team: product, data science, and clinical partners.
Problem
Parkinson's disease is typically diagnosed years after symptoms begin. By the time a formal diagnosis is made, the disease has often been progressing for 3+ years. Early intervention could slow deterioration, but the diagnostic pathway was too slow, expensive, and inaccessible. The business problem was demonstrating that Telefonica's network assets had real clinical value — creating a platform for health partnerships.
My role
I owned the product definition and clinical partner relationship. This included scoping the use case with Hospital Clinic, defining the service model (telephone-based, no app required), setting the requirements for the ML pipeline with the data science team, and determining how clinical outputs would be structured and acted on. Data science model development was delegated to the Telefonica R&D technical team.
Approach
The core product decision was accessibility: making the service work over a standard phone call rather than a smartphone app, which kept it viable for the full patient population regardless of device or age. I defined the clinical framing — outputs as screening flags, not diagnoses — which was the right position given 78% accuracy and shaped how Hospital Clinic integrated results into their workflows. I worked with data scientists on the input requirements and with clinical staff on what a clinically useful output actually looked like in practice.
Outcome
- Model achieved 78% detection accuracy — clinically meaningful for a screening context
- No specialist equipment or smartphone required — accessible across age and geography
- Project demonstrated Telefonica's network data as a health asset, underpinning subsequent health partnership strategy
The product was shelved pending a pharmaceutical treatment for Parkinson's — no cure meant limited clinical demand for early detection at the time.
