Gatik Arena AI simulation
Focus and efficieny in AI simulation process
ROLE
LEAD PRODUCT DESIGNER
STATUS
LAUNCHED/+20 CORE USERS ADOPTION
TEAM
1 DESIGNER, 2 FE, 1 BE
SCOPE
0 > 1 PRODUCT
PROJECT TYPE
30–40% reduction in tracking/review time
One user can confidently handle 50+ jobs without the system breaking
CONSTRAINTS
Engineers' bias towards the swivel chair and sturdy products. Embedding scalability foundations in the MVP
WHAT IS ARENA?
Next-generation simulation platform producing photorealistic, structured, and controllable synthetic data for autonomous vehicle training and validation
GOAL
To design a minimal, scalable, and high-signal dashboard for autonomous vehicle (AV) engineers, focused on efficiently managing Job Creation (Scenario Editing) and Failure Analysis across large simulation test campaigns.
TARGET USER GROUPS
The main user groups were identified as Simulation Engineers and AV Algorithm Developers using desktop computers for long hours, with a high workload in which each item requires attention and analysis.
PROBLEM
Slow autonomy scenario iteration
Creating and modifying scenarios is a bottleneck, especially for edge cases
SOLUTION
Scenario editor that is flexible
Component-based, structured editor for rapid environment and actor manipulation.
PROBLEM
Low signal from simulation results
Creating and modifying scenarios is a bottleneck, especially for edge cases
SOLUTION
Dashboard that highlights failure and success
High-level view prioritizing failed jobs based on severity and traceability.
PROBLEM
Difficulty tracking progress and regression
Hard to know if the latest code base is better than the last.
SOLUTION
Regression & KPI View
Automated comparison of key metrics between software versions.
THE DILEMMA OF TOOLS AND EFFICIENCY
Understanding the users' workflow, problems, and workarounds of Simulation Engineers and AV Algorithm Developers using desktop computers for prolonged hours, with a high workload in which each item requires attention and analysis. The main areas that value could be delivered were efficiency and precision.
DESIGN DEVELOPMENT ITERATIVELY WITH THE STAKEHOLDERS
Recognizing that adoption depended on engineering trust, I brought engineers and developers into early focus-group conversations to align on short-term needs, long-term goals, and real technical constraints. Through this co-design process, we created a tool that solved their most pressing problems and was intentionally designed to scale with future demands.
ITERATION AND EARLY FEEDBACK ON KEY USERFLOWS
FINAL DESIGN DECISIONS
Create simulation jobs that are well-tagged and easy to find
Enable users to modify and create and supervise variants of scenarios with efficiency
Increase team communication by visibility and shareability
Make simulation results as intuitive as possible
IMPACT
Reduced tracking & review time by ~30–40%
Engineers spent considerable amount of time to review the results not included the amount of time spent to find the right simulations.
50+ jobs can be handled confidently by one user reducing the amount of misatkes
The iterative design unlocked the need for the system not to break in usability when users repsobility scales to beyond 50 simulation jobs.
100% adoption rate by the target user group
The MVP was adopted by the full research group and used as the primary review surface during experiment cycles.
Replacing painful swivel chairs
The tool replaced ad-hoc comparison workflows and became the team’s default method for reviewing simulation outcomes.
