I’m Shreya, a Lead Mixed Methods UX Researcher.

Fig. 01 · Agentic AI

Agentic AI Workflow

Connecting research to content design with AI agents

Challenge

Research findings reached content designers slowly and unevenly, often after copy decisions were already made.

Approach

I piloted an agentic AI workflow that brought relevant research findings into content design as copy was drafted and reviewed, with researchers checking the agents’ output.

Impact

Content decisions drew on research by default, and the pilot showed the team where agents help and where people still need to review.

Fig. 02 · Prioritization

Data-backed MVP

Feature prioritization with MaxDiff and conjoint

Challenge

With limited engineering bandwidth for the year, the team needed to know which features delivered the most user value.

Approach

I designed a quantitative MaxDiff study to measure feature importance, followed by a conjoint where participants could choose “no preference” to test how strongly the features really differed.

Impact

A clear, data-driven prioritization model that guided roadmap decisions and aligned stakeholders on where to invest.

Fig. 03 · Benchmarking

Usability Benchmarking

Impact measurement with usability benchmarking

Challenge

Redesigning a complex product without standard success metrics like conversion or revenue, and with limited access to user data.

Approach

A mixed methods usability benchmark, combining qualitative insight (friction, mental models) with quantitative metrics (time on task, errors, SUS score).

Impact

A clear, measurable framework for product success, and a culture of data-driven decisions on the team.

• 250users surveyed
• 5fewer clicks on average
• 12%increase in satisfaction

Fig. 04 · Segmentation

User Segmentation

Data-backed personas through user segmentation

Challenge

Teams across the org had each built their own personas, with no single source of truth and little data behind them.

Approach

A large-scale survey pairing profiling questions with behavioral questions, analyzed to find segments that matter for the product.

Impact

Five data-backed segments replaced 13 overlapping personas, giving teams across the org one source of truth to plan and design from.

• 380users surveyed
• 5user segments
• 13redundant personas removed

Fig. 05 · Rapid research

Workflow Creation

Building a rapid research program

Challenge

Product teams were moving faster than traditional research timelines, so decisions were often made before research could weigh in.

Approach

I built a rapid research program at JPMorgan: a standard intake for requests and short-turnaround studies sized to each decision, so teams could get answers within their planning cycles.

Impact

Research became part of fast-moving decisions instead of arriving after them, and teams came to research earlier with their questions.

Fig. 06 · Methods mentorship

Quant Mentorship

Raising the team’s method choices with MaxDiff and conjoint

Challenge

Researchers defaulted to general surveys, even for trade-off and prioritization questions that surveys answer poorly.

Approach

I mentored researchers on MaxDiff and conjoint principles, including when each method fits and how to read the results.

Impact

The team reached for more targeted methods whenever applicable, and prioritization questions got clearer, more defensible answers.