Experience | Junho Choi

Urban
Conga

UX Design
Consultant

Jan 2026 / Jun 2026

Ideated an interactive public installation system to capture and interpret usage patterns, improving behavioral visibility constraints.

Mozilla
Firefox

IT
Consultant

Aug 2025 / Feb 2026

Built a React + Vite performance extension tracking 50+ tabs in real time. Ran user research, translated findings into Figma UI, shipped with a 12-person team.

Asan
Medical

AI Research
Intern

May 2025 / Aug 2025

Built 3D imaging pipelines for HCC classification. Applied SHAP and Grad-CAM to make model decisions legible to clinicians. Placed 7th in a multi-institutional challenge.

Braven

UI/UX
Consultant

Jan 2025 / May 2025

Synthesized survey, benchmarking, and affinity data for Braven with a team of 8 — driving the mobile app iterations that lifted engagement and usability.

Biosolution

Research
Intern

May 2024 / Aug 2024

Designed microRNA assays — PCR, FACS, HPLC — to characterize osteoarthritis targets. Ran cell culture and in-vivo testing; supported the CAR T-cell program.

US 8th
Army

Non-Commissioned
Officer

Sep 2022 / Mar 2024

Led a bilingual 40-soldier platoon. Translated and delivered ACFT training to a 96% pass rate. Awarded the US Army Certificate of Achievement.

UCSF

Telehealth
Coordinator

Aug 2021 / May 2022

Built longitudinal analyses of telehealth utilization, surfacing the workflow bottlenecks that reshaped clinical delivery and resource allocation.

Scroll or click to explore

Jan 2026 / Jun 2026

Urban Conga

UX Design Consultant

Invention Corps of Berkeley · Berkeley, CA

Collaborated with a group of 9 consultants to ideate a data-informed interactive public installation system capable of capturing and interpreting usage patterns of existing installations, with a focus on improving engagement analytics and behavioral visibility.

Defined data-driven design constraints for low-cost, low-overhead sensing approaches, successfully translating qualitative behavioral visibility goals into structured, quantifiable data schemas.

Developed spatial data concepts and 3D configurations using Rhino to align physical installation layouts with inferred data flows, optimizing the system's capacity for systematic measurement of community engagement.

Data AnalyticsRhino 3DUX ResearchSpatial Design

Project Workflow

Phase 01 · Jan – Feb 2026

Public-space behavioral survey

Built and distributed a survey across the team’s networks asking people how they actually use civic squares, parks, and commercial spaces — group size, time of day, length of stay, deal-breakers, and what they’d feel comfortable doing in public. The goal: capture real behavior, not assumptions, before any installation design started.
Phase 02 · Feb 2026

Insight extraction

Each of the 9 consultants was asked to surface 2–3 patterns from the survey responses. The themes that emerged immediately:
- Almost everyone visits public spaces in groups of 3–5 people
- Overcrowding, cleanliness, and noise are the top deal-breakers
- People are uncomfortable with anything they’re unaware of — but much more relaxed once they understand what an installation is doing
- Behavior splits into two distinct modes: quick passers-by (15–30 min) and extended visitors (1+ hour)
Phase 03 · Feb – Mar 2026

Space-type segmentation

The team recognized that one installation concept can’t cover every context. We grouped responses into three space archetypes — each with its own usage signature, mood, and core values:
- Civic spaces — 2 hours, 2–5 people, productive at noon/night, valuing education, focus, and sightseeing
- Park spaces — 30 min–1 hour, relaxation-focused, valuing exercise, stress relief, quiet, and friends
- Commercial spaces — 1–2 hours, productive at noon/night, social and purchase-driven
Phase 04 · Mar 2026

Concept ideation

Each space archetype got its own installation direction. The brainstorm produced a long-list of physical and digital interaction concepts matched to the space’s mood:
- Litter walls, locks, etching/scratching — permanence and trace-leaving in civic spaces
- Infrared sensing & AI cluster detection — measuring how long and how many, depending on context
- Touch / pressure / button voting — e.g. a Starbucks “Strawberry Acái vs. Caramel Macchiato first to 100” in a mall
- Material wear-and-tear over time — the installation visibly changes from use
- Sticker walls / community blurbs, kinetic park installations, racing/climbing micro-games
Phase 05 · Apr 2026

Sensing & data-schema constraints

Translated the qualitative behavioral goals (“measure dwell time,” “count cluster size,” “detect voluntary engagement”) into structured, quantifiable data schemas with low-cost, low-overhead sensing approaches — capacitive touch, IR proximity, simple pressure switches.
Phase 06 · May – Jun 2026

Spatial & 3D layout in Rhino

Modeled candidate installation footprints in Rhino 3D, validating that each layout could host the inferred data flows from phase 5 while staying within the budget and footprint of a real public space. Closed the loop between behavior → sensing → physical form.

Aug 2025 / Feb 2026

Mozilla Firefox

IT Consultant

Invention Corps of Berkeley · Berkeley, CA

Engineered a Firefox performance monitoring extension using React and Vite that tracked, filtered, and sorted 50+ concurrent tabs by memory and CPU usage in real time, helping users identify and manage high-resource processes.

Led user research through surveys to assess Adaptive Performance perceptions, analyzing feedback to translate user needs into feature requirements and UI designs in Figma, collaborating with 12 team members across frontend and backend tracks.

ReactViteFigmaUser Research

Project Workflow

Phase 01 · Aug 2025

Project framing — Adaptive Performance

Started from a hypothesis: most browsers waste battery on idle tabs, and Firefox could differentiate with an Adaptive Performance mode that intelligently throttles based on usage. Defined the open questions we’d need to answer with users: do they care about eco features, what data would convince them, and how much performance loss is acceptable in exchange for battery gain?
Phase 02 · Sep 2025 (Week 1)

Survey question brainstorm

The 12-person team filled a FigJam board with raw interview/survey question candidates — each consultant contributing from their angle. Questions ranged from blunt (“why do you prefer your browser over Firefox?”) to specific (“if you could save 10% battery at the cost of significantly laggier performance, would you?”).
Phase 03 · Sep 2025

Categorize into themes

Grouped the brainstorm into five working buckets that became the spine of the eventual questionnaire:
- Background / Demographics — age, browser of choice, hardware, power-user level
- Eco Qs — environmental motivation, willingness to trade speed for sustainability
- Power Qs — current habits around tab sleeping, low-power modes, memory awareness
- Convenience Qs — 1-click vs. customizable, visibility preferences for feature output
- Extensions — how integrated browser extensions are into the user’s daily flow
Phase 04 · Oct 2025

Survey structuring

Pressure-tested the brainstorm against eight guiding criteria — purpose, target audience, question types, clarity, flow/logic, scalability, time-to-complete, and ethics. The result: a coherent survey that flows from general to specific, mixes qualitative and quantitative items, and stays under a respondent-friendly length.
Phase 05 · Oct – Nov 2025

Final survey assembly & distribution

Consolidated the brainstorm into a clean, multi-section instrument: Background Screening, Eco-Conscious Preferences, Power-user Questions, Extensions, Browser Usage & Experience, and Final Questions. Distributed across the team’s networks; collected enough responses to read both broad signals and edge cases.
Phase 06 · Dec 2025 – Feb 2026

Build — tab-performance extension

Took the survey insights into engineering. Built a Firefox extension in React + Vite that tracks 50+ concurrent tabs in real time and lets users filter and sort by memory and CPU usage — the “visibility” signal the survey identified as the missing piece in current eco-mode UIs. Designed the UI in Figma with the 12-person team split across frontend and backend tracks.

May 2025 / Aug 2025

Asan Medical Center

AI Research Intern

AMC · Seoul

Processed and analyzed high-dimensional clinical datasets in Python and Jupyter to identify key data patterns, informing feature selection and team strategy in a multi-institutional classification challenge to place 7th.

Built automated data preprocessing pipelines for 3D medical imaging and clinical records to streamline data preparation across multiple modalities, reducing manual processing time for the research team.

Implemented interpretability techniques (SHAP, Grad-CAM) to validate model outputs and communicate feature importance to non-technical stakeholders, increasing trust in data-driven clinical predictions.

PythonPyTorch3D U-NetResNet3DMachine LearningAILLM

Jan 2025 / May 2025

Braven

UI/UX Consultant

Invention Corps of Berkeley · Berkeley, CA

Collaborated with a team of 8 consultants to synthesize multi-channel UX research for Braven — including survey data, competitive benchmarking, and affinity mapping — quantifying user pain points to inform a data-driven design strategy.

Systematized mobile app design iterations by validating user needs through rigorous qualitative and quantitative analysis, producing a measurable increase in engagement and interface usability.

UX ResearchAffinity MappingBenchmarkingMobile UIFigma

Project Workflow

Phase 01 · Jan 2025

User outreach & survey distribution

Designed and distributed a multi-question user survey to undergraduates at public four-year universities, sampling across class years, demographics, and device usage. The instrument captured how students discover career events, why they choose virtual over in-person, what they retain, and the friction points before, during, and after attending.
Phase 02 · Feb 2025

Aggregate & visualize the data

Cleaned the raw responses and translated them into bar charts, pie charts, histograms, and heatmaps so the team could read the dataset at a glance. Surfaced majority opinions, outliers, and inconsistencies across the eight-person research team.
- Visualized attendance, device, time-of-day, and content-preference distributions
- Flagged patterns worth investigating (e.g., low ease-of-use ratings, time-constraint cluster)
Phase 03 · Feb 2025

Affinity mapping

Each consultant generated sticky-note insights from their assigned section of the survey. We then clustered them on a shared FigJam board into three themes: Demographics, Motivations, and Pain Points & Needs — exposing the conflicting points where Braven’s assumptions diverged from what students actually wanted.
Phase 04 · Feb – Mar 2025

Synthesize general audience insights

Turned the clusters into a single working audience profile. Two themes dominated:
- Time constraints & convenience were overwhelmingly the reason students chose virtual events at all
- Students don’t know what they want — they want the end result without the effort, and lack the steps to get there
The base persona: a second-year, public-institution, underserved, Zoom-comfortable, social-media-savvy student.
Phase 05 · Mar 2025

Year-by-year deep dive (Part 2)

The general profile was too coarse to drive design, so we split the dataset four ways — first-year, second-year, third-year, fourth-year — and reran the affinity exercise within each cohort. Each consultant was assigned a year and tagged every survey question with where that cohort’s proportions diverged from the average.
Phase 06 · Mar – Apr 2025

Per-year personas

The deep dive produced four sharply different audience profiles — each with its own demographics, motivations, pain points, and needs:
- First-years — part-time employed, interested in networking and exploring industries. Lack of time is the sole barrier; want 1:1 mentor meetings.
- Second-years — more passive, less proactive about finding events. Want personal meetings and chat/email follow-ups; uncomfortable speaking to employers.
- Third-years — busiest cohort, focused on internships. Highest engagement during events; want resume help and intimate 1:1 meetings with employers.
- Fourth-years — largely unemployed, hunting full-time roles. Want direct employer contact, not more information — fewer webinars, more 1:1s.
These four personas became the spine of the mobile app iterations that followed.
Phase 07 · Apr 2025

Low-fidelity wireframes

Translated the four personas into the first set of mobile screens — pure structure, no styling. Sketched out the Student Roadmap, Upcoming Events, Past Videos / Speakers library, and the Home dashboard as gray-and-purple wireframes. Goal here: validate the information architecture and navigation flow with the Braven team before any visual polish.
Phase 08 · May 2025

High-fidelity mockups

Layered Braven’s brand language (the navy / coral / butter-yellow palette) onto the validated wireframes and pushed every screen to production-ready fidelity in Figma. Final set covered the welcome dashboard, Braven Connect speaker library, Events calendar, Career Questionnaire, My Tasks, and the personalised Career Roadmap — all tied back to the persona-driven decisions from earlier phases.

© 2026 Junho Choi