Design a privacy-safe personalization system for Stock

Problem Statement Description

Product context: Adobe is a creative, document, and marketing software company; its products include Creative Cloud, Photoshop, Illustrator, Acrobat, Adobe Express, Firefly, and Experience Cloud.

Adobe Stock serves creators who need to discover, license, and use high-quality assets across creative workflows. In this interview, you are asked to design a privacy-safe personalization system for Stock that can improve asset discovery while respecting user trust, enterprise expectations, and Adobe’s broader commitments around responsible AI and professional creative use.

The core challenge is to personalize recommendations, search ranking, collections, or creative inspiration without over-collecting sensitive data, exposing confidential project context, or creating experiences that feel intrusive. Consider creators working independently as well as teams and enterprise users who may search for campaign assets, brand-specific visuals, templates, videos, vectors, or generative content inside connected Adobe workflows.

You should frame the system as a Technical PM: define the product requirements, data flows, API or service boundaries, privacy and security constraints, model or ranking inputs at an appropriate level, rollout strategy, observability, and trade-offs. The scope should be concrete enough to discuss what is built, how it integrates with Stock and adjacent creative tools, and how teams would know whether it is safe and useful.

The experience should consider:

- Creator workflows for browsing, searching, saving, licensing, and reusing Stock assets across individual and team projects.

- What personalization signals may be useful, what data should be minimized or excluded, and how consent, controls, and transparency should work.

- Requirements for enterprise customers, including account-level policies, admin controls, data isolation, auditability, and compliance expectations.

- System architecture considerations such as profile services, recommendation APIs, event logging, ranking services, caching, and integration with search or creative tools.

- Privacy, security, and responsible AI risks, including sensitive project inference, cross-user leakage, bias, copyright or licensing concerns, and misuse of behavioral data.

- Reliability and performance expectations for search and recommendation surfaces, including latency, fallback behavior, and graceful degradation.

- Measurement and observability, including product metrics, privacy guardrails, model quality checks, experimentation, incident detection, and user feedback loops.

- Rollout approach, including phased launches, opt-in or policy-controlled access, monitoring, rollback criteria, and communication to creators and enterprise stakeholders.

Your goal is to describe a technically credible, privacy-safe personalization system that improves Stock discovery for creators while preserving trust, control, and professional-grade reliability.