Hi, I'm Darya.

I work at the intersection of product strategy, data science, and UX research — turning complex findings into things teams can actually act on.

DomainsPublic Health · Biotech · Pharma · SaaS · AI/ML
Core StrengthsUX Research · Data Science · Product Strategy
About

How I got here.

My path started in public health.

That grew into an interest in research on people and systems. Over time, working in applied research, I wanted to strengthen my analytical rigor with further education. That took me to the University of California, Berkeley's MIDS program, where I built a formal foundation in data science and machine learning.

From there, through varied experience, I grew as a UX researcher, applied data science, machine learning, and AI to research problems, and moved into product direction and strategy, where all three meet.

That grounding in data science and machine learning is what makes me stronger across both research and product. I stay closely tied to AI, applying data science and ML solutions to research problems and product decisions alike, and increasingly using AI tools throughout my work.

The path here
  1. Public Health — San Jose State University

    Undergraduate degree

  2. Data Science — University of California, Berkeley (MIDS)

    Master's degree

  3. UX Research

    Applied research focused on people and systems

  4. Data Science for UX

    Built on my UX research foundation with data science methods, growing beyond it as my data science skill deepened

  5. Product Direction and Strategy

    Where data science, UX, and product strategy meet

Methods

My toolkit.

My background is mixed methods, and I flourish in that range and experience.

This map shows the breadth of my expertise — spanning all four quadrants, from interviews and co-design sessions to A/B testing, NLP and experimentation.

My entire professional experience has been fast, rapid applied research — gained working in small, agile teams and through my consulting background.

The quant tells you what is happening and how much. The qual tells you why, and for whom.

Numbers give scale and confidence; stories give meaning and direction. Qual adds the color to the quant — which is why my work sits on both sides of the map below.

Filter
Qualitative Quantitative Exploratory Evaluative discover size & scope interpret measure & validate Contextual inquiry Participatory co-design Survey design Focus groups Stakeholder synthesis NLP & text analytics Segmentation & clustering Interviews Rapid-cycle studies Conjoint / MaxDiff Usability testing A/B & experimentation Longitudinal outcomes Statistical model Causal inference
← Swipe the map to explore →
Selected work

Three samples of my work.

Each one below: what was at stake, how I approached it, and what actually changed as a result.

01 · Program evaluation

Does universal basic income change outcomes for youth aging out of foster care?

View the program ↗
Longitudinal survey design Instrument calibration Quantitative outcome analysis Digital storytelling Stakeholder coordination Qualtrics
The problem

Former foster youth face disproportionate risk of homelessness and financial instability once they age out of the system. A government agency needed evidence on whether a universal basic income of $1,000/month for two years measurably changed housing, education, and employment outcomes, in order to decide whether to expand the program. Eligible participants had been in the foster care system between ages 16 and 21 and were living in high-cost Santa Clara County.

My approach
  • Involved from project initiation, coordinating needs across the agency and partner organizations
  • Defined scope and programmed a pre/post design: baseline and matched follow-up surveys, n=72
  • Built and launched the instruments in Qualtrics, using standardized questionnaires so results were comparable across time and to outside benchmarks
  • Owned all quantitative analysis across housing, education, employment, and financial-security domains
  • Co-created digital stories with participants (video, poetry, photography) to show the program's impact — bringing the user lens to the quantitative data
Method & instrument design

I worked on the program's first pilot phase, an initial cohort of 72 foster youth enrolled in 2020. We conducted monthly check-ins with the same 72 respondents, from enrollment through follow-up roughly two years later, tracking eight quantitative outcome measures over time. Surveys were built and fielded in Qualtrics using standardized questionnaires.

Because the participants were current and former foster youth, I calibrated the instruments before launch — testing that every question was understood as intended, and that respondents would interpret and answer it the same way. Comparing change over two years only works if the question means the same thing in month one as it does in month twenty-four, and the same thing to every respondent. That calibration is what made the numbers below trustworthy.

What the data showed — baseline → follow-up
0%→0%
pre
post
Full-time school enrollment
0%→0%
pre
post
Literal homelessness
0%→0%
pre
post
Full-time employment
0%→0%
pre
post
Home ownership
0%→0%
pre
post
Holding a checking account
0%→0%
pre
post
In a self-reported meaningful career
Impact Findings contributed directly to the agency's decision to extend the program and fund it at a larger scale — a second cohort began payments the following year, and the model has since expanded into a nine-pilot regional universal-basic-income effort.
02 · Research program redesign

Building a repeatable research model that kept HR leaders ahead of the trends that mattered.

View a sample study ↗
Research program design Survey research Interviews Rapid-cycle studies Thought leadership
The problem

The organization's research process was built for long, infrequent reports — by the time a study shipped, the workforce and compensation trends it covered had often already moved on. Leadership needed research HR leaders could act on while a trend was still current, not months later. The real problem wasn't any one study; it was that there was no repeatable way to produce them at speed.

My approach
  • Designed a repeatable research model — a standard three-week cycle, question set, interview structure, and output format the team could run again and again
  • Built reusable components: a defined survey question framework, a standing panel of HR professionals, an interview guide template, and an infographic format that stayed consistent study to study
  • Ran the first cycles end to end myself — the research, the analysis, the writing, and the design — to prove the model worked before handing it off
From kickoff to published insight
3 wks
Total time, initiation to published results
30 min
Focused practitioner interviews, clearly defined questions
0
Days to build the infographic once data closed
1
Week 1

Defined clear survey questions and an interview guide; fielded to our existing panel of HR professionals

2
Week 2

Ran focused 30-minute practitioner interviews alongside the survey, using clearly defined questions

3
Week 3

Synthesized findings and built the infographic — published just days after data collection closed

Impact The faster survey-plus-interview cadence let the organization publish credible research while a trend was still forming rather than after the fact — establishing its research team as a go-to source HR leaders turned to for what was current in their space, and building a repeatable model the organization continued to run studies on.
03 · Technical product management, ML/AI

Making a marketing-mix model usable for the people who act on it.

Technical PM Machine Learning Usability testing Statistical model
The problem

The client needed a single source of truth for marketing ROI across every channel — statistically rigorous enough for Data Science and Finance to trust, but interpretable enough for marketing teams to actually use in planning decisions. The platform estimates incremental sales impact, ROI, and marginal ROI at the channel level, with more granular breakdowns available underneath.

My approach
  • Served as the vendor-side product lead — responsible for the platform's data science foundation, building the product end to end, and supporting the client's roadmap and decision-making
  • Ran usability sessions directly with marketing-team users to understand how they actually worked with the platform
  • Reviewed outputs from both a statistical and a user perspective, surfacing where results didn't match how people needed to act on them
  • Guided decisions on methodology, features, and the level of detail exposed — balancing rigor with usability
The model

Built on Google Meridian, an open-source Bayesian marketing-mix model.

Input Marketing & business data

Spend and performance context across channels.

Engine Google Meridian

A Bayesian model reasons about the data probabilistically.

Output Impact & ROI by channel

Incremental sales impact, ROI, and marginal ROI.

My contribution: I helped validate the model's results, built out the features the client requested, and proposed additional features that shaped the roadmap. I also helped develop an optimizer that let marketing teams plan spending scenarios ahead of time — supporting the reallocation of $500M in marketing spend toward higher-performing channels.

Platform at scale
0
Pharmaceutical brands supported by the platform
1Q
Modeling turnaround — fresh enough for marketers to act on, not months stale
Live
Optimizer let marketing teams test planned spend scenarios in real time before committing budget
Key finding & impact The optimizer supported reallocating $500M in marketing spend toward higher-performing channels. From there, I kept expanding the platform — more features, more granular data — so it could deliver faster, more actionable insight for both technical and non-technical users, soliciting ongoing feedback alongside formal usability studies throughout.
Public work

Published research & projects.