QASIM MALIK
UX RESEARCH

I investigate the problem, shape the product, and build enough to learn what happens next.

UX Researcher and product builder working across AI, technical, and emerging products. I use research, design, and rapid prototyping to turn ambiguity into product direction.

SELECTED WORK

DATABRICKS

AGENTIC AI / 0→1 UXR

2026

Established research for a new agentic AI experience, shaping product strategy and interaction models for trust, control, and adoption.

FORMATIVE RESEARCH / MOBILE

2025

Helped adapt a key entry-level advertising experience from desktop to mobile and converge three competing product directions into one.

AMTRAK

SERVICE DESIGN / BUSINESS STRATEGY

2024

Used frontline field research to reframe historic-station modernization and guide executive prioritization of a new sustainability initiative.

ABOUT ME

I’ve worked across AI, advertising, transportation, and enterprise products at Databricks, LinkedIn, Amtrak, and consulting teams supporting companies like Meta, Wikipedia, The Metropolitan Museum of Art, and Civian. Prior to that, I worked in academic and healthcare research, improving digital health technologies, systems, and treatments.

My background spans cognitive science and HCI, and I often work beyond traditional research — using prototyping and AI-assisted tools to explore product ideas and close the gap with design and engineering.

M.S. Human–Computer Interaction, Pratt Institute · B.S. Cognitive Science, University of Connecticut

Outside work, I’m usually reading, taking photos with my far too large collection of early-2000s CCD sensor digicams, or looking for something sweet to eat.

READING NOW

Água Viva - Clarice Lispector

CURRENTLY SHOOTING

Olympus C-4040Z

SWEET OBSESSION

Hellenika Ube & Coconut ice cream

Resume

01 / Case study

CustomerLake

UX Research Intern

2026

Shaping an agentic AI experience from ambiguity to a clearer product direction.

I was the primary researcher on an early-stage agentic AI initiative for marketing and data practitioners. Across three rounds of generative and evaluative research, I helped the team clarify how people should work with an AI agent and where trust, control, and governance needed to shape the experience.

Between studies, I worked with design and product to turn findings into interactive concepts, using prototyping and AI-assisted development to make emerging directions concrete enough to test.

⧉ Introducing CustomerLake: The Agentic CDP embedded in Databricks

Role

UX Research Intern

Timeline

May–August 2026

What I did

User interviews · Concept testing · Co-creation · Interactive prototyping

Partners

Design · Product · Engineering

02 / Product context

CustomerLake brings agentic AI into the customer-data workflow.

CustomerLake is Databricks’ Agentic Customer Data Platform, built natively in Databricks. It brings core CDP capabilities like Customer 360, identity resolution, audience building, campaign automation, activation, and personalization into the same governed data and AI environment enterprises already use.

A central part of the product vision is giving marketers agentic interfaces to work with trusted customer context, while data teams retain governance over the underlying data and models. This introduces a new collaboration model between marketing and data teams.

Where I came in

I joined while that experience was still taking shape. The product direction was ambitious, but important interaction questions were still open:

How should marketers express intent to an agent?

What should the agent do on its own?

What would users need to inspect, verify, or approve?

How could the experience reduce operational friction without weakening governance?

My role was to help reduce that uncertainty through iterative research, then carry the learning forward into product directions we could test.

Initial FRAMING

What would need to be true for users to adopt and trust this kind of AI-assisted workflow?

03 / How the work evolved

Each round narrowed the next product question.

Test the emerging direction

Concept testing + interviews

Question

Which capabilities resonate, and what could block adoption?

Shift

Promising capabilities were not yet adding up to a coherent workflow.

Investigate the riskiest assumption

Generative interviews + workflow research

Question

Would greater self-service actually solve the underlying problem?

Shift

The technical task was not the real bottleneck.

Make the findings tangible

Co-creation + concept development + prototyping

Question

What would the research imply if we turned it into an interaction?

Shift

Research principles became product hypotheses (concepts) we could test.

Test the emerging model

Workflow research + concept testing

Question

Where should AI assist, what do users need to verify, and where should they remain in control?

Shift

A clearer interaction model emerged around translation, verification, control, and governance.

04 / Initial Findings

Powerful AI capabilities weren’t enough.

The product needed to serve people with very different levels of technical fluency. Marketers thought in business goals and campaign outcomes; technical counterparts (operations, data teams) were responsible for making those goals executable, governed, and safe.

Three questions kept surfacing:

Does it fit how we work?

Can we trust it?

Can we actually work this way?

The opportunity became:

How might an AI agent make complex data work more accessible to non-technical users without losing the controls that make it trustworthy?

Translation tension

Business intent

→

AI / agent

→

Governed technical logic

Turning point 01

The first prototypes had features, but not yet a workflow.

Participants saw value in individual capabilities, but the experience did not yet map cleanly to how they worked. The research pointed toward greater guidance, room to iterate, human review around consequential actions, and better visibility into what the agent was doing.

From

Which AI capabilities matter?

How should these capabilities work together as an experience?

Connecting the workflow

Goal

→

Build

→

Review

→

Act

→

Learn

Turning point 02

We were optimizing the wrong bottleneck.

The team was exploring greater self-service as a way to make a slow workflow faster; specifically, granting marketers ability to create a particular data piece needed for building campaigns. Research showed that the technical build itself represented only a small portion of the delay. Most friction came from requirements translation, cross-team coordination, handoffs, queueing, and governance.

From

How can we make creation faster?

How can we compress the translation + governance loop?

Automating execution alone would make one step faster without fixing the system around it.

Current workflow + Bottleneck

01 · Ambiguous goal

Request

02 · Intent → requirements

Translate

03 · People + systems

Coordinate

04 · Technical step

Build

05 · Evidence + risk

Validate

06 · Accountability

Approve

Translation + coordination

Most elapsed work

Build

Small step

Validation + governance

Necessary friction

The build wasn’t the bottleneck.

The opportunity was a tighter loop between business intent, technical interpretation, evidence, and accountable action, not simply faster execution.

Translate intent

Turn ambiguous business language into structured logic.

Make correctness visible

Help users inspect what the system understood.

Embed review

Make approval and accountability part of the experience.

Scale autonomy with risk

Give users more freedom during exploration than during consequential actions.

05 / From insight to interaction

I turned the research into a shared product question.

Rather than handing off a set of recommendations, I brought the findings into a cross-functional co-creation session with design, product, and engineering. We explored what the research meant in the context of the broader workflow, then translated the strongest directions into two contrasting interaction models for testing.

Research findings

Evidence from the workflow

Cross-functional co-creation

Design · Product · Engineering

Two interaction hypotheses

Contrasting models of trust

Structured

Conversational

Prototype + test

Compare the models in research

↔

Finding

Business intent gets lost when it has to become precise system logic.

Product question

How might the agent help with that translation without making its reasoning opaque?

Interaction hypothesis

Make interpretation visible, editable, and verifiable.

Question for testing

Which model better helps users understand, verify, and confidently proceed?

Starting from the team’s existing concept, I forked the direction, independently redesigned the interaction model, and vibe-coded this alternative prototype myself.

01 · Structured interaction

01 · Goal

02 · Build

03 · Verify

04 · Constraints

05 · Proceed

Step 03 · Review the agent’s interpretation

Is this what you meant?

Goal

Editable interpretation

Evidence

Visible source signal

Constraint

Review required

Edit

Confirm interpretation

Break an opaque interaction into inspectable decisions.

Make correction and verification part of the primary path.

Explicit stages make the agent’s interpretation inspectable.

This interaction model was developed collaboratively with the designer as a contrasting hypothesis for testing.

02 · Conversational interaction

Abstracted dialogue sequence

User

Agent · Clarify

User

Agent · Proposed interpretation

Evidence + assumptions surfaced

Dialogue lets intent and interpretation evolve together.

Product details are abstracted to protect confidential work.

Turning point 03

The deeper workflow was translation and judgment.

The final research round zoomed back out. The actual job began before someone manipulated controls: teams clarified the business goal, negotiated what their data could support, worked around incomplete information, validated risk, and decided whether it was safe to act.

The opportunity: help users turn an ambiguous goal and imperfect data into a defensible decision.

AI could help with

Drafting · translating · suggesting · explaining · checking

People needed to retain

Judgment · approval · consequential decisions

Emerging human–AI model

Intent

Translation

Evidence

Review

Action

AI assistance

Draft · translate · reason · check

Human accountability

Increasing consequence

06 / What the work converged on

A clearer interaction model for agentic AI.

Translation

Show how user intent becomes system logic.

The agent needs to expose what it believes the user means.

Verification

Give users enough evidence to decide whether an output is credible.

Confidence comes from intermediate checks, not merely polished outputs.

Control

Make editing, rejection, approval, and recovery normal parts of the interaction.

Collaboration requires meaningful intervention points.

Governance

Increase autonomy only when the consequences allow it.

The appropriate amount of AI autonomy changes with the cost of being wrong.

07 / Impact

Research insights shaped product decisions:

roadmap items touched

roadmap bets validated / strengthened

design changes in-flight

15+

cross-functional stakeholders engaged

Product direction

Helped shift isolated capabilities toward a more coherent workflow.

Roadmap

Strengthened and redirected product bets.

Design

Research and concept testing informed active interaction changes.

Team learning

Co-creation and prototyping sustained an ongoing product-learning loop.

08 / Reflection

What I’m taking forward:

The question can evolve with the research

As the research progressed, the most useful product questions often became clearer too. I learned not to treat the initial framing as fixed, and to treat research as a continuous loop of discovery.

Product building: making can extend research

This project expanded how I see my role as a researcher. Turning findings into prototypes extended the research loop and closed the gap with design and engineering. Making became a way to carry insights further into the product process.

Bringing people in creates momentum

Co-creation gave my team space to engage, share perspectives, and build on the findings together. Collaboration like this makes reserach into something teams can actually move forward with.

Next — LinkedIn / Mobile Boosting →

01 / Case study

Mobile Boosting

UX Research Intern

2025

Bringing LinkedIn’s entry-level advertising experience to mobile.

I led formative research for LinkedIn’s first mobile Boosting experience, helping the team understand how mobile-first SMB advertisers make promotion decisions and evaluate three competing product directions.

3 → 1

Design directions converged

Design changes across three iterations

$35M+

Estimated incremental revenue opportunity de-risked

02 / Opportunity

Boosting worked on desktop. But the audience was increasingly mobile.

Boost allows members to pay to extend an organic post beyond their network. More than half of new advertiser acquisition came through Boost, while many SMB advertisers operated mobile-first.

80%

of LinkedIn sessions were mobile

40%

of weekly posts were mobile

$35M+

estimated opportunity to bring Boost to mobile

03 / Product decision

Three fundamentally different ways to translate Boosting to mobile.

Single

Desktop-like, information-dense long scroll.

Stepped

Focused flow with fewer decisions per screen.

Simple

Preset packages with estimated outcomes responding to selections.

What does a genuinely mobile-native Boosting experience need to do differently?

04 / Research questions

What changes on mobile?

How do mobile context, behaviors, and needs affect design and feature prioritization?

Which direction?

How do advertisers respond to the three interaction models?

What creates confidence?

What information proves a Boost is worth the investment?

05 / Approach

Generative exploration with evaluative concept testing.

moderated remote sessions

60 min

interview + concept testing

design directions

SMB

advertisers with Boost experience

The team had limited understanding of the mobile user group but already had concepts to evaluate, so the study combined contextual interviews with think-aloud prototype testing.

Collaboration

Product · Design · Data Science · Engineering · Research Ops

06 / Process

Recruit → Observe → Synthesize + Recommend

Recruit

Partnered with Data Science on participant identification and Research Ops on screening and scheduling.

Observe

Invited stakeholders to sessions, note-taking, and emerging-findings updates.

Synthesize

Moved from findings to insights, takeaways, recommendations, topline, and full shareout.

07 / Finding 01

Advertisers couldn’t configure a Boost confidently without understanding what their choices would produce.

INPUTS

Audience + Goal + Budget + Duration

→

EXPECTED RESULT

Estimated reach and impressions respond to configuration choices.

→

CONFIDENCE

Is this configuration worth the investment?

Product implication

Show expected outcomes while users configure the Boost by dynamically updating estimated results as variables change.

08 / Finding 02

More control wasn’t necessarily more empowering.

“I’m not a marketer… if there was some kind of AI advisor that could give me: 'people have done this,' or the reason you might want a longer duration, or do this audience — that’d be cool.”

Product implication

Turn setup from a configuration form into a guided decision-making experience: recommendations, smart defaults, and contextual assistance.

09 / Finding 03

Mobile changed the acceptable level of complexity.

Reduce cognitive load

Less content and scrolling at once.

Reduce typing

Favor tap-to-select, steppers, and lightweight controls.

Reduce simultaneous decisions

Focus attention on one configuration choice at a time.

10 / Finding 04

The mobile opportunity didn’t end when the Boost launched.

Respond

React to engagement while momentum is high.

Track

Know when performance reaches meaningful milestones.

Re-engage

Quickly Boost again when something is working.

11 / Direction

Three directions became one.

3 EARLY CONCEPTS

Single · Stepped · Simple

→

RESEARCH

Needs, evidence, tradeoffs, recommendations

→

1 REFINED DIRECTION

Understandable, guided, mobile-native, lifecycle-aware

Make outcomes understandable. Guide instead of overwhelm. Design for mobile behavior. Support the full Boost lifecycle.

12 / Impact

Research changed the product direction.

3 → 1

Convergence from divergent directions to one refined concept.

Design changes across three iterations.

MVP

Priority features shaped the roadmap and MVP PRD.

$35M+

Estimated incremental opportunity de-risked

13 / Reflection

Research isn’t neutral facilitation. It requires a point of view.

I entered a project with three plausible product directions rather than a blank canvas. My role wasn't to just ask users to choose a design. It was to dig into what lay underneath — what information, guidance, control, and feedback made the experience valuable — and use that evidence to recommend where the product needed to go.

Research, at its best, helps teams choose a direction with a clearer understanding of what will make it work. Doing that well requires owning your insights and being opinionated about what the evidence supports.

Next — Amtrak / Historic Stations of the Future →

01 / Case study

Historic Stations of the Future

UX Research Intern

2024

Using frontline research to shape sustainable modernization at Amtrak.

I led ethnographic research inside a working historic station, translating frontline needs into opportunity areas that informed an enterprise-wide innovation challenge and executive prioritization.

2 days

Embedded fieldwork

5 roles

Observed and interviewed

9 opportunities

Translated from themes

1 initiative

Prioritized for development

02 / Opportunity

How do you modernize a historic station without erasing what makes it historic?

Amtrak serves more than 500 stations, roughly one-third of which are listed on the National Register of Historic Places. These buildings hold significant cultural value, but preservation requirements complicate the infrastructure updates needed to support modern operations and Amtrak’s Net-Zero by 2045 strategy.

500+

stations served

~1/3

listed as historic

2045

Net-Zero target

Strategic question

How might Amtrak modernize its historic stations sustainably while preserving their cultural identity?

03 / Reframing the problem

Not just an infrastructure problem: a system to improve.

The original challenge focused on buildings and sustainability. But frontline work crossed physical spaces, digital systems, customer interactions, and informal team dependencies. Improving the station required understanding how all of these elements worked together, not simply identifying new technology to install.

Initial framing

What sustainable upgrades could Amtrak introduce?

→

Research framing

What must sustainable modernization improve for the people operating the station?

04 / Approach

Discovery → Participatory Design → Prioritization

Rather than moving directly into solution generation, the project connected frontline evidence to enterprise decision-making.

Discover

Study frontline work and identify operational friction, unmet needs, and sustainability opportunities.

Translate

Turn observations into themes, role profiles, and actionable opportunity areas.

Prioritize

Use the research to guide employee ideas and evaluate directions for desirability, viability, feasibility, and impact.

05 / Field research

To understand a historic station, I needed to understand the people who keep it running.

I coordinated two days of ethnographic research at Joseph R. Biden Jr. Station in Wilmington, Delaware. Together, five interconnected roles revealed how customer service, station information, safety, baggage operations, and facilities depend on one another.

PIDS operator

Customer service representative

Baggageman

Amtrak police officer

Extraboard employee

06 / Research under real conditions

Research inside an active train station meant being dynamic and adaptable.

Employees were serving customers, monitoring trains, and responding to unexpected events while participating. I structured the study around the Jobs to Be Done framework, then classified questions as essential or optional so each session could be flexible without losing the most important knowledge gaps.

Goals

Tasks and indicators of success

Dependencies

Tools, information, and team handoffs

Friction

Operational and environmental pain points

Context

Historic status and sustainability behaviors

07 / Synthesis

Nine findings converged into three opportunity areas.

Instead of treating each finding as an isolated issue, I organized the evidence into three strategic areas that leadership and employees could act on.

Preserve the station as a cultural asset

Historic character was something customers valued, not merely a constraint on renovation.

Modernize the operational foundation

Wayfinding, aging infrastructure, and internal systems created recurring friction.

Enable sustainable action

Employees wanted to contribute but lacked clear goals, visible progress, and practical tools.

Historic interior architecture at Wilmington Station

08 / The central tension

The station’s history created value and friction at the same time.

Historic architecture gave the station a distinctive identity and attracted public interest. But its layout, infrastructure, and preservation constraints also contributed to customer confusion and operational burden. The opportunity was not to choose between preservation and modernization—it was to make them reinforce one another.

Wayfinding

Poor signage generated frequent employee interruptions.

Infrastructure

Aging HVAC, ventilation, and plumbing affected comfort and safety.

Systems

Outdated tools and information gaps increased cognitive load.

Sustainability

Paper-heavy processes and inefficient energy use persisted in routine work.

09 / Translating research

Beyond a report: the research became a decision-making tool.

I translated findings into opportunity statements that connected observed problems to possible interventions. These gave employees and leadership a shared foundation for imagining solutions grounded in frontline reality.

Observation

Customers struggled with wayfinding and repeatedly interrupted employees for assistance.

→

Opportunity

Improve station orientation and signage so customers can navigate independently while employees remain focused on critical work.

10 / Participatory design

Frontline insights became an enterprise-wide design challenge.

Amtrak invited employees across the organization to propose ways of modernizing historic stations sustainably. The research gave participants and evaluators a shared view of problems occurring on the ground. I also helped two frontline employees translate ideas surfaced during fieldwork into formal submissions.

The strongest proposals advanced from open submission into a structured evaluation process.

Frontline evidence

Observed work and unmet needs

Opportunity areas

Research-backed strategic territories

Employee ideas

Enterprise participation

Executive prioritization

Evidence-informed selection

11 / PRIORITIZATION

Research became the lens for choosing what moved forward.

An Innovation Advisory Committee narrowed the submissions to three finalists. My research report gave evaluators evidence about the most pressing needs on the ground, helping them assess ideas against both frontline reality and organizational constraints.

Desirability

Does it solve an important human problem?

Viability

Does it make sense for the business?

Feasibility

Can Amtrak realistically deliver it?

Impact

Would it meaningfully improve the system?

12 / Outcome

The selected direction made sustainability visible and actionable.

The winning concept was Power Save Mode: a digital platform designed to help stations understand progress toward Amtrak’s sustainability goals and access practical resources for improvement. It directly addressed a major finding: employees were motivated to act sustainably but lacked clear goals, shared definitions, and visible feedback.

Selected direction

Amtrak’s Sustainability team selected the concept as its next large-scale initiative, with Innovation supporting development through research and iterative design.

13 / Impact

Research connected frontline reality to enterprise strategy.

The work grounded sustainability decisions in operating realities and created reusable research infrastructure for future modernization efforts.

1 initiative

Power Save Mode selected for development

25% participation

of idea-platform visitors submitted proposals — above industry average

5 role profiles

created as reusable research assets for future work

9 themes

translated into opportunity areas

14 / Reflection

The highest-leverage research impact was confident alignment.

This project reinforced that research can create impact before a product exists. By connecting frontline work, organizational goals, and evaluation criteria, I helped the team move from a broad question about future stations to a direction grounded in the realities of operating them.

Design research around the environment

While a researcher's curiosity is endless, the time you have with users is not. Flexible sessions and prioritized knowledge gaps preserved research intent without asking employees to step outside their environments.

Rapport is key for quality research

Meaningful research participation depends on trust. Rapport and transparency about how ideas would be used helped employees contribute more candidly and constructively, especially in a context where corporate hierarchies and social dynamics created obstacles.

Connect needs to business decisions

Insights gain influence when they are framed connected to the decisions leaders face. Framing employee experiences against feasibility, organizational priorities, and long-term service outcomes made them easier to act on.

Next — Databricks / CustomerLake →

New York, NY

/in/qmal

qasimhmalik1@gmail.com

(860) 993-5148

Resume

Qasim Malik

Assembling sheet deck