Designing systems that strengthen judgment.
I turn ambiguous problems into products teams can defend. I use research to find the smallest change that matters most, then carry that reasoning through strategy, design, and implementation.
When AI is part of the solution, I decide what the system should contribute, what people should continue to judge, and what should never be automated.
An evolving body of work
Five chapters in an ongoing investigation.
Each project began with a different problem. Research revealed that the hardest constraint was rarely technical. It was protecting the human element behind a successful outcome. The challenge was to augment human capability, not automate it away. These projects show how evidence redirects product direction.
01 StoryJam, collaborative product
Protect independent thinking before the group begins to converge
Traditional brainstorming tools reveal ideas as they arrive, letting early responses shape what follows. I led StoryJam around a different model: independent contribution before simultaneous reveal, reducing anchoring before the group converges.
Read the case study →
02 RefWorks, enterprise SaaS
Expose what changed so people can reconcile the result
Systematic-review teams maintained an audit trail outside RefWorks and needed the product to match it. I reframed the migration problem around reconciliation, then designed workflows that made count changes, duplicate decisions, and project boundaries easier to inspect.
Read the case study →
03 Sensory Sprout, conversational AI
Preserve meal reliability before expanding choice
Sensory Sprout began as a recipe-discovery concept. Research showed that families first needed to protect the small set of meals everyone could reliably eat. I reframed the product around meal reliability, separated safety rules from AI judgment, and made household preferences reviewable.
Read the case study →
04 Mezzo, ambient product
Support awareness without demanding attention
Mezzo began as a real-time volume display. Working software showed that feedback failed when receiving it became another task. I reframed it as an ambient companion and simplified its language, states, and responsibilities so awareness could happen at a glance.
Read the case study →
05 Prototypes as discovery, applied AI
Delay the intervention until the point of leverage is clear
Two working AI prototypes appeared useful in isolation. Workflow testing revealed their real value was diagnostic: they exposed what had to change before AI could help. I reframed them as instruments for finding the point of leverage.
Read the case study →
Approach
Design the system around the judgment the outcome depends on.
I treat the initial request as a starting point. Through research, I identify the human capacity the outcome depends on, find the smallest change that matters most, and choose an intervention that strengthens rather than erodes it.
Practice
Strengthen the human capacity the outcome depends on
I identify what people still need to do well and design greater capability to support it.
Method
Find the point of leverage before choosing the intervention
I examine workflows, constraints, incentives, risks, and human needs to understand what produces the current result and where a focused change could matter most.
AI Specialization
Decide what the system should handle and what should remain human
When AI is part of the product, I define what it can contribute reliably, what people must continue to judge, and what should not be automated. The goal is a responsible division of work, not maximum automation.
Evidence + Delivery
Let evidence redirect the work and preserve the reasoning through delivery
Research should change the problem, not merely validate the solution. I connect strategy, design, engineering, AI behavior, and implementation so the reasoning survives from discovery to a working product.