dMetrics

At dMetrics, I helped redefine Minsky AI around a more fundamental product question: how should people work with machine-learning results when the useful insight lives in the connections between information, evidence, and entities? I carried that question from strategy and desk research into user flows, interaction models, prototypes, interface design, and testing.

Minsky AI

A data-first interaction model for machine-learning workflows

The product could not make complex analysis useful by presenting isolated model outputs. Users needed to explore the underlying information, understand relationships, and decide what to act on without losing the technical context behind a result.

• Reframed the product around connected information. I helped shift the direction from consuming static outputs to working with entities, evidence, and the relationships between them.
• Turned machine-learning concepts into usable interactions. I designed flows and interface patterns that made model capabilities understandable, controllable, and actionable without hiding their underlying complexity.
• Explored the system before committing to screens. Desk research, journey mapping, and prototyping helped define how users would enter the data, move between related information, inspect evidence, and evaluate results.
• Tested the interaction model with users. I used feedback from prototypes and platform evaluation to identify where the concepts or controls were unclear, then revised the flows before implementation.
• Worked across the technical stack. I partnered with front-end and back-end engineers and ML/data scientists to reconcile user needs with model behavior, data structures, and implementation constraints.

The work changed the product’s organizing model, not only its interface. It established the data-first foundation that later evolved into Minsky’s knowledge-graph-oriented, evidence-grounded analytical workflows.