Kiwi EV: Full redesign of LingOS 2.0
01 Background: A Broken Foundation
The 2023 Kiwi EV promised Gen-Z a "fun, trendy" urban driving experience. However, user satisfaction surveys showed users were deeply frustrated with the Kiwi EV's in-car system, LingOS1.0, and it was hurting sales. But SAIC-GM couldn't pinpoint why or how to fix it.
So they asked us to redesign the entire OS in 4 months without changing the fixed 10.25-inch hardware. Typically, this scope requires 12+ months.
LingOS 1.0
The Core Conflict
How to deliver a safe and delightful interaction system for drivers & co-drivers within extreme time and physical constraints?
LingOS 2.0
02 Building Order: Creating Structure Before Design
When I took ownership of the project, the biggest challenge wasn't the design itself—it was organizational chaos.
The Real Constraint:
- No complete PRD
- No system-level prototype
- Fragmented UI assets with untraceable decision history
In practice, the organization had lost a "shared understanding" of the OS.
Reconstructing the Truth
To untangle the technical debt, I led the team in reverse-engineering the entire OS by rebuilding:
- A complete feature inventory
- The system-level information architecture (IA) and wireframes
- Functional dependencies and Layering across the OS
Result:
- Exposed structural redundancies clearly
- Became the single "source of truth" for alignment with PM and engineering
This became the foundation of OS 2.0.
Navigating the Stakeholder Maze
Very quickly, another constraint surfaced:
Marketing, Product, and Engineering each had conflicting, scattered—yet reasonable—demands.
Under a 4-month deadline, unresolved conflict would have derailed the project. To prevent this, I established a Decision Governance Framework at the beginning:
- The North Star Principles
- Conflict Filtering Matrix: User value V.S. Deliverability within 4 months
This transformed subjective preferences into measurable trade-offs.
03 Design Strategy: Pruning for Performance
A system-level HMI audit revealed that OS 1.0 failed not because of execution quality, but because it ignored the driving context.
Core Issues:
- Critical actions placed in visual blind zones
- Redundant overlap between software and physical controls
- Multi-layer UI structures competing for space on a 10.25-inch display
I addressed this by applying a pruning strategy: removing redundancies and refocusing the system around core functions.
Solution 1
The Dock: Ergo-Centered Reform
In-Vehicle Ergonomic Testing
Problem:
- In driving task tests, icon scan time increased exponentially with icon count. Error tap rate spiked significantly beyond 4 icons
- Left-side Dock area suffered from steering wheel occlusion (15% of screen area)
- Software functions duplicated physical buttons (e.g., climate control)
Insight:In the driving context, the Dock isn't a "feature hub"—it's a safety-critical control zone. Extra icons only increase visual search cost.
The Decision:
- Reduced icons from 8 → 4
- Position change. I proposed two solutions:
Proposal A: Center alignment
Proposal B: Radical move
Proposal A was adopted using Conflict Filtering Matrix.
Impact:
- Shorter visual dwell time
- Larger tap target zones
- Dock transformed from "feature collection" to "safety component"
Solution 2
Music: Winning Back Vertical Space
LingOS 1.0 Music
VOC+ Hierarchical Card Sorting
Music is one of the highest-frequency use cases in the system, yet OS 1.0 structurally undermined it(VOC and card sorting).
Problem
- Music tasks had the longest visual dwell time. Root cause is that vertical space compressed by Dock + Mini Player layers, resulting in extremely low information density
- The biggest complain from both driver and co-driver is: can’t search and switch the songs easily.(VOC)
Decision:
- Integrate Mini Player into the global Dock
- Removed card-style content presentation
New layout: Mini Player in Dock + Vertical List
04 Balancing Innovation & Business: The Dual Desktop
The KIWI's dual desktop concept – a widget desktop and a Card desktop – was a non-negotiable product requirement. Instead of challenging the constraint itself, I focused on resolving the added interaction complexity through innovative interaction design.
LingOS 1.0 Dual Desktop
Problem:
- Users couldn't clearly distinguish functional boundaries between desktops
- Accidental desktop switches occurred frequently
Insight:The problem isn't "whether dual desktops should exist," but whether they have clear mental roles.
Decision:
Redefining Desktop Mindsets
Card Desktop - Emotional scenarios
- Introduced "Scenario Store" with cards like Camping, KTV
- Reserved interface for future service expansion
Widget Desktop - Maximum efficiency
- Merged all apps for daily high-frequency needs
- Optimized for quick access during driving
The Switching Solution:
Remove the biggest point of friction
The original design used bottom swipe + dock icon to switch desktops. User research revealed two critical issues:
- Too precise for driving: Small touch targets made blind operation difficult and unsafe
- Inconsistent behavior: Users couldn't predict which method to use, breaking the dual-desktop mental model
Initial approach: 4-finger pinch gesture
- Easier to trigger without pinpoint accuracy
- Larger gesture area improved success rate
User testing revealed: Steep learning curve for new users
Final implementation: Three-tiered system
- 4-finger pinch
- Voice command: "Switch to card desktop"
- Auto-switch logic: System suggests desktop type based on vehicle status (architecture prepared for future implementation)
*4. Onboarding tutorial
Impact:
- Eliminated accidental switches
- Reserved scalable interface for future services
- Maintained marketing differentiation while ensuring safety
05 Visual Philosophy: The "Glance"
Introducing a Light Theme wasn't purely aesthetic. In the complex lighting conditions of a vehicle cabin, a high-contrast light theme paired with minimal UI significantly improved glance-readability, ensuring users can instantly capture information at high speeds.
Inspired by LIGHT, we designed Icons, Voice AI Assistant, Time Card by time changing.
Intelligent Driving Mode Illustration. Partial designed by UI designers
Accessibility Testing with Multiple Color Palettes and UI Variations.
06 Validation: Physical Buck Testing
Due to the 4-month development cycle not allowing multiple iterations on production vehicles, we partnered with Unity to build the Physical Buck (cabin simulator) for high-frequency prototype validation, differentiating parked vs. driving contexts.
Testing Protocol:
- 30 participants (mix of internal staff and target users)
- Scenarios: Highway driving, urban traffic, nighttime conditions
HMI UX Measurement Framework
Buck Testing
Key Discoveries & Iterations:
Issue 1: Initial Dock icon spacing insufficient, 18% mistap rate → Fix: Adjusted to minimum 48px touch zones, mistap rate dropped to 3%
Issue 2: Light theme caused severe glare at night → Fix: Introduced adaptive brightness system, automatically reducing white ratio at night
Issue 3: Music list scroll speed too fast, users couldn't stop precisely → Fix: Added inertial damping algorithm, scroll experience satisfaction improved 40%
Final result: Completed 3 high-frequency iteration rounds within 4 weeks.
56%
Reduction in average total glance time per task
32%
Reduction in average task completion time
37%
Increase of user satisfaction score
07 Scaling
Design System Deliverables:
- 1,000+ pages of components, guidelines, and documentation
- Comprehensive button & icon library
- Color system with day/night mode support
- Supported by multiple UI designer contractors
08 Impact:
Post-Lauch Adoption
Business Impact:
- Scaled across 5 SAIC-GM vehicle models (2023-2025)
- 3-year lifespan as SAIC-GM's OS standard
- Design system adopted by 3 subsequent vehicle programs, saving 6 months of development time per program
Reflection
What I learned:
As a senior designer, my value wasn't just delivering a UI—it was establishing a rational decision-making logic within organizational chaos, making design the core driver of product delivery.
The broader lesson:Systematic knowledge accumulation accelerates product design speed and efficiency. The UX measurement framework we built wasn't just used in this project—it became an adaptive tool for the entire organization. Due to time constraints, we didn't conduct the complete measurement suite, but extracted core function tests that proved the framework's flexibility.
Kiwi EV: Full redesign of LingOS 2.0
01 Background: A Broken Foundation
The 2023 Kiwi EV promised Gen-Z a "fun, trendy" urban driving experience. However, user satisfaction surveys showed users were deeply frustrated with the Kiwi EV's in-car system, LingOS1.0, and it was hurting sales. But SAIC-GM couldn't pinpoint why or how to fix it.
So they asked us to redesign the entire OS in 4 months without changing the fixed 10.25-inch hardware. Typically, this scope requires 12+ months.
LingOS 1.0
The Core Conflict
How to deliver a safe and delightful interaction system for drivers & co-drivers within extreme time and physical constraints?
LingOS 2.0
02 Building Order: Creating Structure Before Design
When I took ownership of the project, the biggest challenge wasn't the design itself—it was organizational chaos.
The Real Constraint:
- No complete PRD
- No system-level prototype
- Fragmented UI assets with untraceable decision history
In practice, the organization had lost a "shared understanding" of the OS.
Reconstructing the Truth
To untangle the technical debt, I led the team in reverse-engineering the entire OS by rebuilding:
- A complete feature inventory
- The system-level information architecture (IA) and wireframes
- Functional dependencies and Layering across the OS
Result:
- Exposed structural redundancies clearly
- Became the single "source of truth" for alignment with PM and engineering
This became the foundation of OS 2.0.
Navigating the Stakeholder Maze
Very quickly, another constraint surfaced:
Marketing, Product, and Engineering each had conflicting, scattered—yet reasonable—demands.
Under a 4-month deadline, unresolved conflict would have derailed the project. To prevent this, I established a Decision Governance Framework at the beginning:
- The North Star Principles
- Conflict Filtering Matrix: User value V.S. Deliverability within 4 months
This transformed subjective preferences into measurable trade-offs.
03 Design Strategy: Pruning for Performance
A system-level HMI audit revealed that OS 1.0 failed not because of execution quality, but because it ignored the driving context.
Core Issues:
- Critical actions placed in visual blind zones
- Redundant overlap between software and physical controls
- Multi-layer UI structures competing for space on a 10.25-inch display
I addressed this by applying a pruning strategy: removing redundancies and refocusing the system around core functions.
Solution 1
The Dock: Ergo-Centered Reform
In-Vehicle Ergonomic Testing
Problem:
- In driving task tests, icon scan time increased exponentially with icon count. Error tap rate spiked significantly beyond 4 icons
- Left-side Dock area suffered from steering wheel occlusion (15% of screen area)
- Software functions duplicated physical buttons (e.g., climate control)
Insight:In the driving context, the Dock isn't a "feature hub"—it's a safety-critical control zone. Extra icons only increase visual search cost.
The Decision:
- Reduced icons from 8 → 4
- Position change. I proposed two solutions:
Proposal A: Center alignment
Proposal B: Radical move
Proposal A was adopted using Conflict Filtering Matrix.
Impact:
- Shorter visual dwell time
- Larger tap target zones
- Dock transformed from "feature collection" to "safety component"
Solution 2
Music: Winning Back Vertical Space
LingOS 1.0 Music
VOC+ Hierarchical Card Sorting
Music is one of the highest-frequency use cases in the system, yet OS 1.0 structurally undermined it(VOC and card sorting).
Problem
- Music tasks had the longest visual dwell time. Root cause is that vertical space compressed by Dock + Mini Player layers, resulting in extremely low information density
- The biggest complain from both driver and co-driver is: can’t search and switch the songs easily.(VOC)
Decision:
- Integrate Mini Player into the global Dock
- Removed card-style content presentation
New layout: Mini Player in Dock + Vertical List
04 Balancing Innovation & Business: The Dual Desktop
The KIWI's dual desktop concept – a widget desktop and a Card desktop – was a non-negotiable product requirement. Instead of challenging the constraint itself, I focused on resolving the added interaction complexity through innovative interaction design.
LingOS 1.0 Dual Desktop
Problem:
- Users couldn't clearly distinguish functional boundaries between desktops
- Accidental desktop switches occurred frequently
Insight:The problem isn't "whether dual desktops should exist," but whether they have clear mental roles.
Decision:
Redefining Desktop Mindsets
Card Desktop - Emotional scenarios
- Introduced "Scenario Store" with cards like Camping, KTV
- Reserved interface for future service expansion
Widget Desktop - Maximum efficiency
- Merged all apps for daily high-frequency needs
- Optimized for quick access during driving
The Switching Solution:
Remove the biggest point of friction
The original design used bottom swipe + dock icon to switch desktops. User research revealed two critical issues:
- Too precise for driving: Small touch targets made blind operation difficult and unsafe
- Inconsistent behavior: Users couldn't predict which method to use, breaking the dual-desktop mental model
Initial approach: 4-finger pinch gesture
- Easier to trigger without pinpoint accuracy
- Larger gesture area improved success rate
User testing revealed: Steep learning curve for new users
Final implementation: Three-tiered system
- 4-finger pinch
- Voice command: "Switch to card desktop"
- Auto-switch logic: System suggests desktop type based on vehicle status (architecture prepared for future implementation)
*4. Onboarding tutorial
Impact:
- Eliminated accidental switches
- Reserved scalable interface for future services
- Maintained marketing differentiation while ensuring safety
05 Visual Philosophy: The "Glance"
Introducing a Light Theme wasn't purely aesthetic. In the complex lighting conditions of a vehicle cabin, a high-contrast light theme paired with minimal UI significantly improved glance-readability, ensuring users can instantly capture information at high speeds.
Inspired by LIGHT, we designed Icons, Voice AI Assistant, Time Card by time changing.
Intelligent Driving Mode Illustration. Partial designed by UI designers
Accessibility Testing with Multiple Color Palettes and UI Variations.
06 Validation: Physical Buck Testing
Due to the 4-month development cycle not allowing multiple iterations on production vehicles, we partnered with Unity to build the Physical Buck (cabin simulator) for high-frequency prototype validation, differentiating parked vs. driving contexts.
Testing Protocol:
- 30 participants (mix of internal staff and target users)
- Scenarios: Highway driving, urban traffic, nighttime conditions
HMI UX Measurement Framework
Buck Testing
Key Discoveries & Iterations:
Issue 1: Initial Dock icon spacing insufficient, 18% mistap rate → Fix: Adjusted to minimum 48px touch zones, mistap rate dropped to 3%
Issue 2: Light theme caused severe glare at night → Fix: Introduced adaptive brightness system, automatically reducing white ratio at night
Issue 3: Music list scroll speed too fast, users couldn't stop precisely → Fix: Added inertial damping algorithm, scroll experience satisfaction improved 40%
Final result: Completed 3 high-frequency iteration rounds within 4 weeks.
56%
Reduction in average total glance time per task
32%
Reduction in average task completion time
37%
Increase of user satisfaction score
07 Scaling
Design System Deliverables:
- 1,000+ pages of components, guidelines, and documentation
- Comprehensive button & icon library
- Color system with day/night mode support
- Supported by multiple UI designer contractors
08 Impact:
Post-Lauch Adoption
Business Impact:
- Scaled across 5 SAIC-GM vehicle models (2023-2025)
- 3-year lifespan as SAIC-GM's OS standard
- Design system adopted by 3 subsequent vehicle programs, saving 6 months of development time per program
Reflection
What I learned:
As a senior designer, my value wasn't just delivering a UI—it was establishing a rational decision-making logic within organizational chaos, making design the core driver of product delivery.
The broader lesson:Systematic knowledge accumulation accelerates product design speed and efficiency. The UX measurement framework we built wasn't just used in this project—it became an adaptive tool for the entire organization. Due to time constraints, we didn't conduct the complete measurement suite, but extracted core function tests that proved the framework's flexibility.
Kiwi EV: Full redesign of LingOS 2.0
01 Background: A Broken Foundation
The 2023 Kiwi EV promised Gen-Z a "fun, trendy" urban driving experience. However, user satisfaction surveys showed users were deeply frustrated with the Kiwi EV's in-car system, LingOS1.0, and it was hurting sales. But SAIC-GM couldn't pinpoint why or how to fix it.
So they asked us to redesign the entire OS in 4 months without changing the fixed 10.25-inch hardware. Typically, this scope requires 12+ months.
LingOS 1.0
The Core Conflict
How to deliver a safe and delightful interaction system for drivers & co-drivers within extreme time and physical constraints?
LingOS 2.0
02 Building Order: Creating Structure Before Design
When I took ownership of the project, the biggest challenge wasn't the design itself—it was organizational chaos.
The Real Constraint:
- No complete PRD
- No system-level prototype
- Fragmented UI assets with untraceable decision history
In practice, the organization had lost a "shared understanding" of the OS.
Reconstructing the Truth
To untangle the technical debt, I led the team in reverse-engineering the entire OS by rebuilding:
- A complete feature inventory
- The system-level information architecture (IA) and wireframes
- Functional dependencies and Layering across the OS
Result:
- Exposed structural redundancies clearly
- Became the single "source of truth" for alignment with PM and engineering
This became the foundation of OS 2.0.
Navigating the Stakeholder Maze
Very quickly, another constraint surfaced:
Marketing, Product, and Engineering each had conflicting, scattered—yet reasonable—demands.
Under a 4-month deadline, unresolved conflict would have derailed the project. To prevent this, I established a Decision Governance Framework at the beginning:
- The North Star Principles
- Conflict Filtering Matrix: User value V.S. Deliverability within 4 months
This transformed subjective preferences into measurable trade-offs.
03 Design Strategy: Pruning for Performance
A system-level HMI audit revealed that OS 1.0 failed not because of execution quality, but because it ignored the driving context.
Core Issues:
- Critical actions placed in visual blind zones
- Redundant overlap between software and physical controls
- Multi-layer UI structures competing for space on a 10.25-inch display
I addressed this by applying a pruning strategy: removing redundancies and refocusing the system around core functions.
Solution 1
The Dock: Ergo-Centered Reform
In-Vehicle Ergonomic Testing
Problem:
- In driving task tests, icon scan time increased exponentially with icon count. Error tap rate spiked significantly beyond 4 icons
- Left-side Dock area suffered from steering wheel occlusion (15% of screen area)
- Software functions duplicated physical buttons (e.g., climate control)
Insight:In the driving context, the Dock isn't a "feature hub"—it's a safety-critical control zone. Extra icons only increase visual search cost.
The Decision:
- Reduced icons from 8 → 4
- Position change. I proposed two solutions:
Proposal A: Center alignment
Proposal B: Radical move
Proposal A was adopted using Conflict Filtering Matrix.
Impact:
- Shorter visual dwell time
- Larger tap target zones
- Dock transformed from "feature collection" to "safety component"
Solution 2
Music: Winning Back Vertical Space
LingOS 1.0 Music
VOC+ Hierarchical Card Sorting
Music is one of the highest-frequency use cases in the system, yet OS 1.0 structurally undermined it(VOC and card sorting).
Problem
- Music tasks had the longest visual dwell time. Root cause is that vertical space compressed by Dock + Mini Player layers, resulting in extremely low information density
- The biggest complain from both driver and co-driver is: can’t search and switch the songs easily.(VOC)
Insight:Mini Player's essence isn't "content"—it's system status. It shouldn't compete with the content area for space.
Decision:
- Integrate Mini Player into the global Dock
- Removed card-style content presentation
Impact:
- Released 30% vertical space
- Applied intuitive vertical infinite scrolling
- Significantly reduced visual search and attention-switching costs
Measured Result:Music task completion time: 8.2 seconds → 4.9 seconds (40% reduction)
- 1.0 path: Open Music → Horizontal scroll → Tap song = avg. 4.2 interactions
- 2.0 path: Direct Dock access + Vertical list = avg. 2.5 interactions
New layout: Mini Player in Dock + Vertical List
04 Balancing Innovation & Business: The Dual Desktop
The KIWI's dual desktop concept – a widget desktop and a Card desktop – was a non-negotiable product requirement. Instead of challenging the constraint itself, I focused on resolving the added interaction complexity through innovative interaction design.
LingOS 1.0 Dual Desktop
Problem:
- Users couldn't clearly distinguish functional boundaries between desktops
- Accidental desktop switches occurred frequently
Insight:The problem isn't "whether dual desktops should exist," but whether they have clear mental roles.
Decision:
Redefining Desktop Mindsets
Widget Desktop - Maximum efficiency
- Merged all apps for daily high-frequency needs
- Optimized for quick access during driving
Card Desktop - Emotional scenarios
- Introduced "Scenario Store" with cards like Camping, KTV
- Reserved interface for future service expansion
The Switching Solution:
Remove the biggest point of friction
The original design used bottom swipe + dock icon to switch desktops. User research revealed two critical issues:
- Too precise for driving: Small touch targets made blind operation difficult and unsafe
- Inconsistent behavior: Users couldn't predict which method to use, breaking the dual-desktop mental model
Initial approach: 4-finger pinch gesture
- Easier to trigger without pinpoint accuracy
- Larger gesture area improved success rate
User testing revealed: Steep learning curve for new users
Final implementation: Three-tiered system
- 4-finger pinch
- Voice command: "Switch to card desktop"
- Auto-switch logic: System suggests desktop type based on vehicle status (architecture prepared for future implementation)
*4. Onboarding tutorial
Impact:
- Eliminated accidental switches
- Reserved scalable interface for future services
- Maintained marketing differentiation while ensuring safety
05 Visual Philosophy: The "Glance"
Introducing a Light Theme wasn't purely aesthetic. In the complex lighting conditions of a vehicle cabin, a high-contrast light theme paired with minimal UI significantly improved glance-readability, ensuring users can instantly capture information at high speeds.
Inspired by LIGHT, we designed Icons, Voice AI Assistant, Time Card by time changing.
Intelligent Driving Mode Illustration. Partial designed by UI designers
Accessibility Testing with Multiple Color Palettes and UI Variations.
06 Validation: Physical Buck Testing
Due to the 4-month development cycle not allowing multiple iterations on production vehicles, we partnered with Unity to build the Physical Buck (cabin simulator) for high-frequency prototype validation, differentiating parked vs. driving contexts.
Testing Protocol:
- 30 participants (mix of internal staff and target users)
- Scenarios: Highway driving, urban traffic, nighttime conditions
HMI UX Measurement Framework
Buck Testing
Key Discoveries & Iterations:
Issue 1: Initial Dock icon spacing insufficient, 18% mistap rate → Fix: Adjusted to minimum 48px touch zones, mistap rate dropped to 3%
Issue 2: Light theme caused severe glare at night → Fix: Introduced adaptive brightness system, automatically reducing white ratio at night
Issue 3: Music list scroll speed too fast, users couldn't stop precisely → Fix: Added inertial damping algorithm, scroll experience satisfaction improved 40%
Final result: Completed 3 high-frequency iteration rounds within 4 weeks.
56%
Reduction in average total glance time per task
32%
Reduction in average task completion time
37%
Increase of user satisfaction score
07 Scaling
Design System Deliverables:
- 1,000+ pages of components, guidelines, and documentation
- Comprehensive button & icon library
- Color system with day/night mode support
- Supported by multiple UI designer contractors
08 Impact:
Post-Lauch Adoption
Business Impact:
- Scaled across 5 SAIC-GM vehicle models (2023-2025)
- 3-year lifespan as SAIC-GM's OS standard
- Design system adopted by 3 subsequent vehicle programs, saving 6 months of development time per program
Reflection
What I learned:
As a senior designer, my value wasn't just delivering a UI—it was establishing a rational decision-making logic within organizational chaos, making design the core driver of product delivery.
The broader lesson:Systematic knowledge accumulation accelerates product design speed and efficiency. The UX measurement framework we built wasn't just used in this project—it became an adaptive tool for the entire organization. Due to time constraints, we didn't conduct the complete measurement suite, but extracted core function tests that proved the framework's flexibility.