Kavya Jyothi

Kavya JyothiKavya JyothiKavya Jyothi
  • Home
  • My works
  • More
    • Home
    • My works

Kavya Jyothi

Kavya JyothiKavya JyothiKavya Jyothi
  • Home
  • My works

AI-Powered Troubleshooting Portal

A full UX case study showcasing research to design

Overview

Role: Senior UX Designer (Lead)Domain: Industrial UX + AI Systems (Semiconductor Manufacturing)
Users: Fab Technicians, Remote Service Engineers, SMEs, Managers
Deliverables:  Research → Personas → Affinity Map → IA → Flows → Wireframes → UI → Prototype → Testing

Duration: Approx. 6 months
Goal: Reduce troubleshooting time and empower technicians to resolve eligible issues quickly using AI + a safe Quick Fix action.

Context & Background

High-volume semiconductor equipment consists of complex controllers, hundreds of sensors, actuators, and interlocks.
When abnormalities occur, fab technicians file service requests. Engineers must analyze logs, identify root causes, and guide fixes.

The existing process was:


  • Slow (40–80 minutes per issue)
  • Manual log interpretation
  • Inconsistent troubleshooting 
  • Repeated simple issues consuming engineer time 
  • No real-time explainable AI support 
  • Limited visibility for technicians

User Research

Stakeholder Interviews

  • Engineering managers
  • Product owners 
  • AI/ML team 
  • SMEs 
  • Process engineers

User Interviews

  • 12 remote service engineers
  • 6 fab technicians 
  • 3 maintenance managers

Contextual Inquiry - Observed:

  • Controller log interpretation 
  • Parameter cross-checking 
  • Trend analysis 
  • Ticket lifecycle

Artifact Review

  • 200+ sample log files 
  • Error code tables 
  • SOPs & maintenance manuals 
  • Past service requests

User Persona

Persona 1 - Fab Engineer “Daniel”


  • Age: 32  
  • Experience: 4 years  
  • Motivation: Increase production, reduce dependency on SMEs  
  • Pain Points: Logs take too long to analyze and too many repeated issues
  • Needs: Troubleshooting guide, quick issue resolution

Empathy Map

Says

  • “I need to find the root cause quickly.”
  • “Why use multiple tools for one issue?”
  • “These minor problems keep coming back.”
  • “I wish the logs were clearer.”  

Thinks

  • “There should be a faster way to analyze logs.”
  • “I want to handle common issues without SMEs.”
  • “Misdiagnosing affects tool uptime.”
  • “The system should show root causes clearly.”  

Does

  • Switches between logs, dashboards, and docs.
  • Cross-checks alerts and timestamps manually. 
  • Contacts SMEs when stuck or unclear. 
  • Tracks recurring issues and potential fixes.  

Feels

  • Frustrated by repeated low-severity issues. 
  • Stressed during urgent alerts. 
  • Overloaded managing multiple tools.
  • Motivated when solving issues independently.  

Hears

  • “Can you avoid escalating this?” 
  • “Logs are too hard to read.” 
  • “We need quicker turnaround.”  

Sees

  • Disconnected tools and fragmented data.
  • Cluttered logs with technical noise.
  • Recurring issues with no lasting solutions.
  • Pressure to reduce downtime and escalations.  

Affinity Mapping

I synthesized all research into 5 major categories:


Need of AI

Users want clear, understandable predictions.


Address knowlege gap

Engineers want troubleshooting guide to follow.


Need of real time tool tracking

Real time tool monitoring was required


Automatic log collection

Logs are slow to gather and it was difficult to capture right log.


Issue 'Quick fix'

27% of issues are simple and repeat—can be solved with one-click Quick Fix.

Ideation

I conducted structured ideation for concept development using:


Crazy 8s

Generated 8 variations for:

  • AI Insights 
  • Quick Fix UI 
  • Error state handling 
  • Step guidance
     

How Might We…”

  • HMW automate simple fixes? 
  • HMW make AI explainable? 
  • HMW standardize troubleshooting? 
  • HMW support technicians without risk?
     

Co-Design Sessions (with SMEs)

Reviewed:

  • Which issues can be safely automated 
  • Which actions technicians can perform 
  • Required safety validations

Information architecture

  • Dashboard with real-time tool status
  • Create and manage tickets 
  • Quick Fix for eligible recurring issues 
  • Step-by-step troubleshooting guidance 
  • Ticket activity timeline and update 
  • AI-driven insights for faster diagnosis 
  • Controller health snapshot 
  • Standardized troubleshooting flows 
  • Integrated knowledge base 
  • Access to past resolutions and history

Wireframes

Low-Fi wireframe

  • Created rough sketches for Quick Fix screen
  • Log insight module 
  • Root-cause dashboards 
  • Ticket overview.

Mid-Fi wireframe

  • Data density planning
  • Hierarchy planning
  • Navigation model 
  • Step-by-step flow layouts

High-Fi wireframe

  • Industrial dark mode + light mode
  • Final components
  • Visual anomaly graphs
  • Interlock warnings
  • Progress states

User Interface Design

I created interactive prototypes in Figma for:

  • AI Insight Dashboard 
  • Quick Fix Flow 
  • Troubleshooting Stepper 
  • Ticket Timeline 
  • Controller Health Summary
     

Included:

  • Interaction states 
  • Edge cases 
  • Error screens 
  • Success/failure validation


Due to confidentiality, I am not displaying the final UI. At a high level, this is how the first version of the interface looked. The product is currently sold to customers at a high price and remains confidential.

Usability Testing

Participants

  • Multiple fab engineers 
  • Service technicians
     

Tasks Tested

  • Using Quick Fix 
  • Understanding AI suggestions 
  • Navigating troubleshooting flows 
  • Reading anomaly graphs 
  • Completing resolution steps
     

Outcomes

  • Quick Fix highly appreciated 
  • AI became “trustworthy” only after adding explainability 
  • Technicians needed simplified language 
  • Engineers liked dense data view
     

Iterations Based on Findings

  • Added rationale under Quick Fix
  • EImproved trend graph clarity
  • Added technician-friendly vocabulary
  • Used collapsible sections to control density

Impact

Quantitative

  • 27% of tickets resolved using Quick Fix 
  • 42% reduction in troubleshooting time 
  • 30% reduction in manual log analysis 
  • 2× faster onboarding for new engineers
     

Qualitative

  • Engineers felt empowered by structured flows 
  • Technicians gained autonomy 
  • SMEs approved knowledge retention from system

What I learnt

Designed an AI-driven diagnostic and troubleshooting platform for complex semiconductor equipment. Introduced explainable AI predictions, guided troubleshooting workflows, and a safe Quick Fix feature enabling technicians to resolve low-risk issues instantly—reducing ticket load, improving transparency, and significantly accelerating issue resolution.


  • AI in industrial UX must be explainable 
  • “Quick Fix” requires strict safety rules 
  • Technicians value autonomy 
  • Engineers value structure 
  • Domain knowledge grows through SME collaboration 
  • Industrial UX demands clarity, trust, and zero ambiguity

← View Other Projects

This website uses cookies.

We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.

Accept