Understanding how people think
when technology changes.
I'm a UX researcher who enjoys working on products where the answers aren't obvious. Over the past seven years I've helped teams navigate uncertainty through strategic and evaluative research — from early-stage startups to, most recently, Meta, where I worked on products used by millions of people.
How research changed the decision
Some examples of research that shaped what a team chose to build.Click any project to expand.
Which usability problems were actually worth engineering investment?
I led comparative benchmarking research that measured comparable user behaviours across competing products, combining qualitative observation with structured scoring to distinguish cosmetic issues from those that genuinely affected confidence and task success.
Shifted engineering priorities from simplifying workflows to improving predictability and feedback. The benchmarking framework was later reused across subsequent research programmes.
A generalised view of the framework: each issue mapped back to the dimensions it affected, showing not just where a product lagged but why. Illustrative, not actual data.
How should an AI product balance automation with user control?
I evaluated two fundamentally different interaction models using interactive prototypes designed to expose where trust broke down. Rather than comparing usability alone, the research focused on how people reasoned about AI decisions, uncertainty, and control.
Established three design principles that became the foundation for subsequent AI features and informed the team's phased roadmap.
Show the work
AI outputs include visible reasoning. People verify by inspecting, not by trusting.
Communicate failure
Say what the system understood, what it missed, and why. Silent partial results erode trust.
Preserve control
The system proposes; a person decides. Nothing takes effect until it's been approved.
How do you make long-term product decisions when both the technology and the market are changing rapidly?
I synthesised multiple research streams — including qualitative studies, behavioural data, competitive intelligence, and industry trends — into a single strategic framework that helped teams distinguish established evidence from emerging hypotheses.
Aligned multiple teams around a shared AI maturity model and established a trust-first approach for introducing increasing levels of AI autonomy.
Assistive
AI assists; people keep full control. Trust is earned through transparency.
Person works, AI helps
Proactive
AI suggests and automates; people review and approve. Trust is established.
AI works, person approves
Autonomous
Agents work end to end; people set strategy and guardrails. Trust is foundational.
AI orchestrates, person guides
All examples are based on real work but have been anonymised to protect confidential products, metrics, interfaces, and internal strategy.
Beyond individual studies
Alongside product research, I enjoy building the systems that help research scale.
- Created reusable benchmarking frameworks adopted across multiple research programmes.
- Built an AI-powered research repository that enabled teams to self-serve prior research.
- Facilitated cross-functional strategy workshops to build shared understanding before roadmap planning.
- Mentored researchers and collaborated with designers and product managers on research quality and decision-making.
Before Meta, I spent several years helping early-stage teams establish research practices. As the founding UX researcher at an ML startup, I built participant recruitment, research infrastructure, and stakeholder practices from the ground up. Alongside this, I partnered with non-technical founders as a freelance researcher, using generative research to turn early ideas into validated product directions and shippable MVPs.
About
I'm Mari, a UX researcher based in London. Over the years I've lived and worked in Georgia.
Moving between places and disciplines is a big part of how I got comfortable working in unfamiliar territory.
I enjoy working on products at moments of change — when new technologies emerge, user behaviour shifts, or teams need to make decisions without clear precedent.
I'm particularly interested in Human–AI Interaction — not simply how people use AI, but how they decide when to trust it, when to challenge it, and how products can preserve meaningful human agency as systems become more capable.
Meta — AI UX2022–now
Intrro — ML-powered B2B hiring startup2019–21
Freelance — 0 → 1 projects2020–21
MSc Human–Computer Interaction · University of Trento
BA Social Sciences · Free University of Tbilisi