OpenMined — Capstone
Exploring practical product opportunities for privacy-preserving AI through user-centric, ethics-first use-case design.
Context
Privacy-preserving machine learning is technically promising but difficult for non-specialists to adopt. This capstone focused on translating complex privacy concepts into real product use cases that teams can evaluate and build.
Challenge
- Bridge the language gap between technical privacy methods and product decisions.
- Identify use cases where trust and compliance are as important as model accuracy.
- Design concepts that feel usable, transparent, and ethically defensible.
A recurring issue was that privacy is usually communicated as a legal checkbox, not a user experience. The project reframed privacy as a product quality: legible controls, clear data boundaries, and predictable behavior under high-stakes conditions.
Approach & Solution
- Mapped high-risk domains where data sensitivity and trust are mission-critical.
- Evaluated candidate scenarios across desirability, feasibility, and ethical risk.
- Drafted UX concept flows showing how privacy guarantees are made understandable in-product.
The final direction emphasized explainable consent states, explicit sharing boundaries, and low-friction decision points for users. Instead of treating privacy as invisible infrastructure, the concepts made protection and control visible where it matters most.
Outcomes
- Produced a clearer framework for prioritising privacy-preserving AI use cases.
- Aligned technical possibilities with user-facing product opportunities.
- Strengthened my ability to design at the intersection of AI ethics and usability.