Overview

Posture Perfect is an AR-powered fitness application designed for at-home workouts. The product addresses a real gap in the market: guided workout apps exist everywhere, but none offer real-time posture correction feedback — the feature that reduces injury risk and keeps users engaged long-term.

The Problem

Post-pandemic, home fitness shifted from necessity to lifestyle. Yet the apps serving this audience were built around video libraries and static images — with zero feedback on whether users were actually doing movements correctly. The result: users either got injured, felt unmotivated, or both.

"Online workout videos are not tailored as per my needs and goals."

— Research participant

Research

We conducted in-depth user interviews and surveys targeting both regular fitness enthusiasts and people trying to build a home workout habit. Three dominant themes emerged:

User Persona

From our research we developed "Mom Madeline" — a working parent who wants the efficiency of home workouts without sacrificing safety or motivation. Her needs anchored every design decision: convenience, real-time guidance, and gamification elements that kept her coming back.

Design Process

We moved from context scenarios through sketching, wire blocks, and low-to-mid fidelity wireframes — each iteration grounded in Madeline's documented scenarios. The sketching phase focused on four core flows: choosing workouts, performing movements with AR feedback, tracking progress, and motivational elements.

Customer Journey Map — Pre, During, and Post workout phases

Usability Testing & Issue Resolution

We ran structured usability testing with 8 participants across three issue matrix cycles. Each cycle identified problems, rated them by impact and frequency, and drove specific design changes:

Final Designs

The final product delivers a TV-optimized AR fitness experience with a live instruction screen, real-time posture correction overlays, and a wind-down session post-workout. The design system ensures visual consistency across the experience.

Home Screen — Hi-Fi Design

Workout Completion — Wind Down Screen

Design System

Challenges & Learnings

The most significant challenge — and the most valuable learning — was executing a full platform pivot mid-project. Switching from mobile to TV app based on user feedback required rebuilding interaction patterns from scratch. It was uncomfortable. It was also exactly what user-centered design demands: following the evidence, not the original plan.

Coordinating across time zones within the team strengthened our async collaboration habits — a skill that translates directly into distributed product team environments.

Overview

The San Francisco Public Library serves a diverse community with a mission of free and equal access to information. Despite that mission, the website's information architecture was creating real barriers — confusing labels, overwhelming navigation, and content organization that didn't match how users actually think about library services.

The Problem

An initial audit of 26 pages revealed systematic problems: information overload, misleading navigation labels, and an architecture built around internal organizational logic rather than user mental models. The gap between what the library offered and what users could actually find was significant.

Discovering User Needs

I began with a comprehensive exploration of the site across three UX dimensions: user needs, business goals, and content offerings. From this analysis, I compiled 26 common tasks that users would realistically attempt on the SFPL site — from finding books to accessing citizenship services.

Three personas were developed to represent the breadth of SFPL's actual audience: a student researcher, a parent accessing children's programs, and an older adult seeking community resources. Each persona highlighted different navigation priorities and pain points.

User Personas — Student, Parent, Community Member

Tree Testing — Baseline

I conducted a Tree Test using OptimalWorkshop with 10 participants spanning ages 17–67 with varying tech proficiency. Participants completed 10 tasks against the current site structure. The results were revealing:

Card Sorting

To understand users' actual mental models, I ran both in-person and online card sorting. 11 participants total worked with 30 cards representing site content. On average, participants created 8 groups — revealing natural clustering patterns that differed significantly from the existing navigation structure.

Similarity matrix analysis, dendrograms, and best-merge methods revealed consistent patterns in how users grouped content — particularly around task-based organization rather than departmental logic.

Dendrogram — Hierarchical clustering from card sort results

Building the New IA

Card sorting insights drove a new abstract IA and sitemap. I then applied Label Scoring — evaluating each proposed label against five criteria: Well-known, Easy, Real, Brief, and Representative. Labels scoring 0–2 required changes; labels scoring 4–5 were kept. This systematic approach removed guesswork from navigation decisions.

Two navigation options emerged: a Traditional Approach and a Hybrid Approach. I selected the Hybrid Approach for its balance of familiar structure with task-based sub-navigation.

Final Site Map — Post-redesign IA

Tree Testing — Validation

The new IA was tree-tested against the same 10 tasks with the same participant profile. The results validated the approach across every measurable dimension:

Wireframes

With the IA validated, I moved into wireframing — from paper sketches through low-fidelity to high-fidelity mockups. The high-fidelity designs demonstrate the practical impact of the new IA: a cleaner homepage hierarchy, improved search with contextual sub-navigation, and a reorganized Books, Movies & More page with local navigation for quick access.

Homepage — Hi-Fi Wireframe (Improved IA)

Improved Top Navigation — Hybrid Approach

Challenges & Learnings

Starting with a 100% failure rate on one task was jarring — and instructive. It confirmed that good intentions in site structure don't translate to usable navigation without user validation. The iterative approach — test, analyze, restructure, test again — was the only path to a result grounded in real evidence rather than designer assumptions.

Overview

WaitNoMo is a mobile application designed to help busy parents research, shortlist, and enroll their children in preschools — all in one place. The product addresses a deeply stressful life event that currently has no reliable, consolidated digital solution.

The Problem

Preschool enrollment is emotionally high-stakes and logistically chaotic. Parents navigate scattered websites, outdated information, opaque waitlists, and reviews they can't trust. The experience is designed around administrative convenience — not around parents who are time-poor, emotionally invested, and making a decision that affects their child's development.

User Research

I conducted in-depth interviews and surveys with 15 parents. Research focused on the full preschool search journey — from initial awareness through enrollment — capturing emotions, workarounds, and friction points at every stage. These conversations informed the product's core feature set and prioritization.

Research Synthesis — Key Parent Pain Points

Concept Model & Ecosystem Map

A Concept Model visualized how the app's three core features — Search, Waitlist Updates, and Enrollment — interconnected. This guided the design direction and ensured we weren't designing isolated features but an integrated flow.

The Ecosystem Map extended this to identify all touchpoints in the parent's journey, both within and outside the app — school communication systems, email/SMS notifications, and follow-up activities that the app needed to account for at its edges.

Concept Model — Core Feature Relationships

Persona & Journey Map

Maggie — a working mother with limited time — became the design's anchor persona. Her journey map tracked emotional states across searching, waitlisting, and enrollment, revealing the moments of highest anxiety (waitlist submission) and highest relief (confirmation notifications). These peaks and valleys directly shaped the app's notification design and information hierarchy.

Customer Journey Map — Maggie's Preschool Search Experience

Usability Testing

Five moderated usability sessions were conducted using Zoom screen recording. Participants completed core tasks: searching with filters, adding schools to favorites, and checking waitlist notifications. Key findings shaped final design decisions:

"The filters were exactly what I needed — I could narrow it down so fast."

— Usability testing participant

Final Designs

The final mobile interface delivers a consolidated, trustworthy experience across the full enrollment journey — from initial search through waitlist management. Every screen was validated against Maggie's needs and the usability testing findings.

Search Results — School Listings with Real-time Waitlist Data

Map View — Spatial Preschool Discovery

Search Filters — Location, Budget, Hours, Type, Ratings

Challenges & Learnings

The emotional dimension of this project was its biggest design challenge. Preschool enrollment isn't just a logistical problem — it's a trust problem. Parents aren't just finding a school; they're making a decision about their child's safety and development with imperfect information. Designing for that emotional context required extra care in copywriting, information hierarchy, and the way data was surfaced.

The search bar copy change — from "zip code" to "Start the school search" — was a small tweak with outsized impact. It's a reminder that microcopy is UX.