Accelerate Automotive UI Development with Kiro, Squish, and Virtual Targets on AWS

In this post, you’ll learn how AWS and Qt Group collaborated to support innovation in automotive user interface (UI) development. By combining AI-powered code generation (Kiro) and automated testing (Squish) using virtual targets on AWS, customers like Schaeffler, Nissan, Stellantis and others can work to reduce defects and accelerate development workflows.

The automotive industry is transitioning from hardware-centric design to the Software-Defined Vehicle (SDV). This evolution is most visible in the Human-Machine Interface (HMI), which has shifted from being a functional interface to the primary differentiator and a safety-critical component. Modern vehicles integrate multiple high-resolution displays, (e.g., instrument clusters, infotainment systems, and Head Up Displays (HUD) which may require software which complies with ISO 26262 safety standards. The software must also deliver fluid, personalized user experiences. Current development processes, largely reliant on sparse, expensive hardware prototypes and late-stage testing, struggle to manage this complexity and pace. This process tends to defects propagating into customers’ vehicles, leading to growing number of software-related recalls making headlines.

New to Kiro or Squish?

• Kiro is an agentic AI solution for code generation with emphasis on spec-driven development and efficient tools either with an Integrated Development Environment (IDE) or via command line interface
• Squish is a cross-platform graphical user interface test automation tool providing robust, object-based testing for embedded, desktop, mobile and web, with 20+ years of history

By combining Kiro’s AI-powered spec-driven development, Squish’s automated Behavior Driven Development (BDD) testing, and virtual embedded development targets on AWS, development teams can transform how they build automotive user interfaces. Application developers, middleware engineers, UI designers and AI frontier agents all work from the same specifications, seeing results in real-time, leveraging virtual replicas of the devices on AWS.

The Automotive UI Development Challenge

The Automotive HMI space is experiencing robust growth as vehicles increasingly incorporate sophisticated digital experiences. Industry analysts forecast substantial expansion in this sector over the next decade, driven by the proliferation of digital displays in vehicles. There are going to be more, larger and higher performance screens that need to be powered by sophisticated, functionally safe tech stacks along with high performance compute hardware.

Modern HMI development faces convergence of multiple modalities: touch, gesture, voice, Augmented Reality (AR) HUDs, and eye-tracking. AI enables personalization and safety features like fatigue detection, while vehicles connect seamlessly with smartphones and smart home ecosystems. As HMIs adopt safety-critical roles, validation extends beyond functional testing to systemic robustness and safety standards adherence.

A conflict rises in the need for personalization of customer experiences, versus keeping as much stack commonality for cost efficient implementation. These orthogonal demands drive exponential growth of software testing, which traditional setups and workflows struggle to counter. AI-powered automation shifts from being optional to essential for rapid development and testing.

Application developers build user-facing features, middleware engineers handle UI-to-vehicle communication, and platform developers ensure compatibility across Android Automotive OS, QNX, or Linux. Each layer depends on the other layers, with teams traditionally waiting for hardware, including test teams validating on physical targets.

The traditional approaches push testing to the end of the development cycle, when fixing defects becomes exponentially more expensive. Industry research on software defect economics demonstrates that the cost of fixing issues increases exponentially based on when they’re discovered in the development lifecycle[1].

Table 1- Cost impact of late-stage testing

Solution Architecture Overview

Development teams often work across global locations. UI teams may sit in California, middleware developers may be in Germany, and Tier-1 suppliers might be in Japan. The cloud addresses these challenges by enabling customers to have secure multi-account collaboration. This approach maintains IP protection while ensuring seamless integration. It helps transforms traditional development cycles into efficient, parallel workflows. Application developers can implement and validate code using Squish, allowing BDD and integration into continuous integration/continuous delivery (CI/CD) pipelines enabling continuous testing and rapid feedback on every code commit. Developers can run validated code on virtual Android Automotive OS devices instead of physical hardware, using AWS Graviton for instruction set parity. Globally accessible vEDTs (virtual Embedded Development Targets) make end2end testing possible within hours rather than weeks.

Squish offers full support for BDD using Gherkin, while Kiro promotes spec-driven development. With this, Squish and Kiro combine the non-deterministic power of generative AI with the deterministic reliability of traditional test automation. This integration provides test script generation and recording capabilities.

Kiro’s AI-powered, spec-driven development generates initial Squish test scripts from natural language specifications and existing stub test scripts. Squish provides the robust, platform-agnostic testing framework to execute and validate these scripts. Execution results flow back to Kiro via the Squish Model Context Protocol (MCP) server. This direct feedback loop enables Kiro to automatically fix some of the errors in generated test scripts, catching defects early at substantially lower costs when fixing costs in production.

Figure 1- High-level workflow illustrates an automotive UI development automation. Bug Management System and Requirement Management System feed inputs into Kiro, which uses Large Language Models to generate test scripts from natural language specifications. Kiro sends test scripts to Squish via MCP Server for execution on a Virtual Target Fleet running on AWS. Test results and verification reports flow back through MCP Server to Kiro, creating a continuous feedback loop. The system produces verified code and test reports while integrating with CI/CD pipelines for shift-left testing. Through Amazon WorkSpaces, globally distributed teams can securely access Kiro and Squish. They work with vEDTs running on AWS in a workflow centered on synchronized natural language specifications as a single source of truth. vEDTs run the same software stack as physical devices, enabling parallel workflows that catch integration issues early. BMW demonstrated this approach by scaling their virtual testing infrastructure, to reduce cost, accelerate time-to-market and improve our customer’s experience.

Figure 1- High level workflow

What This Means for Your Team

Application developers can write and test UI logic instantly on virtual Android Automotive devices and debug software issues, reducing time delays in waits for physical hardware. Middleware engineers validate service layers against vEDTs, testing error handling and edge cases that teams find difficult to reproduce physically while, integrating with other teams without exposing implementation details. UI and user experience designers see their designs running on virtual displays in real time, iterating animations and interactions immediately, and sharing interactive working prototypes rather than static mockups with stakeholders. Quality assurance engineers scale testing from limited physical devices to hundreds of virtual configurations, running regression suites on every commit and catching integration issues early.

The transformative power lies in early and continuous testing that fundamentally changes the economics of quality. Teams receive feedback in minutes instead of weeks, iterating confidently without breaking existing functionality. Virtual devices with environmental-parity execution show exactly what users will experience. Teams can experience and test touch responsiveness, animation smoothness, visual appearance across themes, and behavior under stress conditions, eliminating surprises like, “it worked in the simulator but feels sluggish on hardware”. This shift-left approach means quality becomes continuous rather than a gate. Everyone shares responsibility for delivering robust, safety-critical automotive interfaces. Application developers see test results in their IDE, while product managers track real-time quality metrics.

Conclusion

The ongoing shift towards SDVs elevates HMIs from functional components to safety-critical market differentiators. Reliance on expensive physical hardware hampers current automotive UI development. The high cost of late-stage defect detection makes this approach unsustainable in the long term.

The architecture demonstrates that Kiro’s AI-powered, spec-driven development together with Squish’s automated BDD testing using vEDTs on AWS, helps accelerate the necessary transformation. This cloud-native approach enables continuous integration and testing by reducing the hardware bottleneck. Development and validation can now occur in parallel from the requirements phase.

This results in a unified, development workflow where teams synchronize natural language requirements with code and tests. Ultimately, this empowers automotive teams to deliver the complex, high-resolution, safety-compliant UIs the market demands, at speed and scale. It enables better quality by catching defects when they’re cheapest to fix.

Ready to get started? Explore Kiro’s AI-powered development capabilities and Squish test automation to begin transforming your automotive UI development workflow. To access ready to use virtual targets visit AWS Marketplace. Learn more at AWS for Automotive, and Qt Group for Automotive.

[1] Tamas Cser. “The Cost of Finding Bugs Later in the SDLC.” Functionalize, January 5, 2023. https://www.functionize.com/blog/the-cost-of-finding-bugs-later-in-the-sdlc; “The Hidden Costs of Firmware Bugs – and How to Avoid Themd.” Developex, https://developex.com/blog/hidden-costs-of-firmware-bugs/.