Eunji Kim | UX & Product Designer Portfolio

eunzoaz@gmail.com

resume

AI-Driven Safety & Fatigue Management for United Airlines

Peer Sky — Designing a fatigue-aware AI system for frontline worker support

AI-Driven Safety & Fatigue Management for United Airlines

Peer Sky — Designing a fatigue-aware AI system for frontline worker support

Team

Five Designers

Timeline

4 Months

Overview

Designed a fatigue intelligence feature within Peer Sky to reduce injuries among United Airlines ramp workers. By combining self-reported and objective data, the system helps managers monitor team fatigue, take AI-assisted actions, and communicate more empathetically. It also streamlines injury reporting and provides safety teams with actionable insights for prevention.

Impact

System Usability Scale (SUS)

Feedback from UA Safety team

My Role

Product Designer

Led end to end design project 0-1

Overview

Design Process

Our process moved from prototypes to usability testing, then to refining designs based on feedback and insights. We repeated this cycle to continuously improve the solution and ensure it met real user needs.

Background

Safety is a top priority at United Airlines, especially for ramp workers. Lately, there’s been a growing focus on fatigue, since it’s increasingly tied to on-the-job injuries.

Ramp worker fatigue is a growing safety risk.

Who’s Ramp Workers?

Ramp workers are ground crew members who work below the wings at airports. They often handle tasks like loading and unloading baggage, guiding aircraft, and operating ground support equipment.

Problem Discovery

Why Do Ramp Workers Experience Fatigue?

A seniority-driven work culture leads to an unequal distribution of tasks, significantly contributing to higher fatigue levels among ramp workers.

We visited the site and spoke with ramp workers to better understand their pain points. Through our observations, we uncovered a work culture driven by seniority, where senior workers are assigned easier tasks while junior workers are burdened with more difficult and less desirable hours.

A seniority-driven work culture leads to an unequal distribution of tasks, significantly contributing to higher fatigue levels among ramp workers.

Seniority-Based Work Hierarchy

Seniority-Based Work Hierarchy

What Systems Exist at the Safety Team and Manager Levels?

To understand how fatigue is being addressed beyond the ramp, we sat in on weekly safety team meetings and observed day-to-day operations on site. From there, we mapped out workflows using a service blueprint, which helped us see how fatigue and injury-related issues are (or aren’t) handled across different roles.

Key challenges at the organizational level:

No Fatigue Data

Complicated Injury Reporting

Lack of prevention strategies

Service Blueprint of Fatigue & Injury Touchpoints

Reframing Problem

Fatigue is a symptom of deeper root causes.

Root Causes → Symptoms

Inefficient, seniority-driven work culture → High fatigue among workers

Lack of standardized fatigue metrics → No fatigue data collected

Complicated injury reporting system → Inaccurate injury data + No effective prevention strategy

Root Causes → Symptoms

Inefficient, seniority-driven work culture → High fatigue among workers

Lack of standardized fatigue metrics → No fatigue data collected

Complicated injury reporting system → Inaccurate injury data + No effective prevention strategy

The Real Problem Lies in a Broken System Where Each Stakeholder Faces Unique Challenges at Different Levels.

Fatigue Challenges Across Stakeholders

Insights to Solution

Designing Features Around Stakeholder Challenges

As we uncovered the underlying problem, it became clear we weren’t just designing for ramp workers, but for a broader network of stakeholders. To explore solutions at different levels, we looked at how other industries tackled similar challenges and applied those insights to our context. We then generated ideas, grouped them to find patterns, and used Excel to prioritize features based on impact and feasibility.

Cross-Industry Research

Affinity Mapping

Excel Prioritization

First, Fatigue needs to be measurable.

To manage fatigue, it first needs to be made visible. Daily self-reports from ramp workers and objective metrics can provide managers with the clarity they need to take meaningful action.

Objective Fatigue Metric

Subjective Self-Reports

Mood Level

Energy Level

Objective Fatigue Metric

Subjective Self-Reports

Mood Level

Energy Level

With the right datasets and clear standards, solutions can operate at different levels—prevent, react, and improve—to proactively manage problems and drive progress.

Insights to Solution

Designing Features Around Stakeholder Challenges

First, Fatigue needs to be measurable.

To manage fatigue, it first needs to be made visible. Daily self-reports from ramp workers and objective metrics can provide managers with the clarity they need to take meaningful action.

With the right datasets and clear standards, solutions can operate at different levels—prevent, react, and improve—to proactively manage problems and drive progress.

Prototype

Wireframe

Iteration

Usability Testing

After developing the wireframes, we invited core users to walk through the designs, participate in usability testing, and share feedback to help identify pain points.

Key insights included:

Users felt uncertain when it wasn’t clear which data was AI-generated versus real.

Non-role-specific information overwhelmed users.

Users struggled to interpret the data and understand what actions to take.

They expected the interface to prioritize information by relevance and usage frequency.

Users feel overwhelmed by too much information

After the first round of usability testing, I no longer had access to core users. To continue improving the design, I gathered quick feedback from 2 to 4 people for each version and explored multiple directions before arriving at the final interface.

The example below shows one of several iterations we refined over time:

v1.0: All data visible, no clear hierarchy, required swiping through team members

v2.0: Click-to-expand list of members with individualized views

v2.1: Added filters, a team-wide status overview, and simplified charts

v2.2: Final version presented data in digestible sections, improving clarity and ease of use

Bob, a ramp worker, is feeling fatigued and has sustained a minor injury while performing his duties. He needs assistance to ensure safety and properly report his condition.

Ramp Worker

USECASE #1

Final Design

Manager Cale receives a report from the Peer Robot regarding Bob’s fatigue and injury. Cale needs to assess Bob’s condition and determine how to adjust the work schedule accordingly to support him.

Manager

USECASE #2

Manager

USECASE #2

The safety team needs to review injury data from all workers to gain a comprehensive view, identify potential patterns, and take corrective actions to prevent future incidents

Safety Team

USECASE #3

eunzoaz@gmail.com

resume

Reflection

How and When AI Should Step In

Working on this project made me think more intentionally about how AI and humans can collaborate smoothly. Through ongoing feedback, I learned how users perceive and respond to AI in real time. The real UX challenge wasn’t what the AI could do, but how and when it should step in. I focused on making handoffs between human and machine feel smooth by setting clear expectations, asking just enough from the user, and offering a chance to preview or edit. Small details in the handoff made a big impact on trust and usability.

There’s still so much to learn, especially as user behavior continues to evolve alongside AI itself.

eunzoaz@gmail.com

resume

eunzoaz@gmail.com

resume

eunzoaz@gmail.com

resume

eunzoaz@gmail.com

resume