case under nda—the visuals are illustrative and the real ui is hidden
scope of work
The platform keeps the product-hypothesis cycle visible from strategic problem to conclusion in one shared workspace.
Problems connect to scored hypotheses, experiments and shared activity. Each experiment records plan, launch, metrics, conclusion and recommendation, with contextual AI support and external metric sources. Teams work in dedicated spaces while discovery remains shared.
I designed the concept, information architecture, four-role flows, UI system, connected content model, hypothesis graph, entity-level AI layer and metric-source architecture.
product decisions
Experiment as a structured object. Plan, launch, metrics, conclusion and recommendation preserve a trace other teams can use later.
Hypotheses form a graph linked to strategic problems, child ideas and neighbouring teams’ experiments, turning isolated work into a connected knowledge network.
Completed experiments enter a tagged repository with metrics and conclusions. Meaning-based search helps teams find relevant work beyond exact keywords.
AI stays in context: it suggests metrics, assesses risk, finds similar experiments and helps turn data into a conclusion.
A shared activity feed gives cross-team visibility, while threaded discussion remains attached to each hypothesis.
Metric-source integrations bring experiment data into the hypothesis, reducing manual chart copying.
Future work explores predictive impact, next-hypothesis recommendations and automatic quarterly reporting from the shared experiment record.
work stages
problem definition Distributed teams repeated experiments and lost conclusions across markets. I framed the recurring issue as a product opportunity.
Quarterly syncs exposed duplicate experiments and lost conclusions across markets. I turned this recurring problem into a product definition.
building the team As early users arrived, a Product Designer, Product Analyst and Frontend Engineer joined. I remained Lead Product Designer and Product Owner.
I built the first working version with Cursor and Windsurf, owning product decisions and architecture while AI assisted with implementation.
active development I lead roadmap, priorities, reviews and metrics as the product expands knowledge, AI search and integrations.
