Service Overview

Unit Test AI

FPT Automotive's unit-test-ai accelerates unit testing via Azure AI and AST analysis. It reduces test case creation from 20-40 min to 5-7 min at 85-90% accuracy, automates MBD rule checks (70% coverage, 20 min saved per model), and integrates with GHA CI, VS Code DevBoxes, and virtualized SILS boards. Establishes a tighter validation loop with centralized test data management under Entra ID governance.

LLMTest GenerationMBDModel Validation

5-7 mins

Test case generation time

85-90%

Test generation accuracy

70%

MBD rule-check coverage

35%

Testing manual effort cut

18.7×

Model validation speedup

Capabilities

Key capabilities

Automated Test Generation

AI generates comprehensive unit tests from source code, reducing manual test writing effort significantly.

Code Coverage Optimization

Intelligent test generation targeting uncovered code paths for maximum coverage.

Digital Cockpit Expertise

Specialized models trained on automotive cockpit software patterns and requirements.

IVI/Cluster/HUD Support

Comprehensive testing for all major cockpit components.

CI/CD Integration

Seamless integration with continuous integration and deployment pipelines.

Technology

Technology stack

Unit Test AI architecture diagram
Component Technology
AI Model LLM, GPT-based
Testing Framework JUnit, CppUnit, GoogleTest
Code Analysis Static analysis tools
CI/CD Jenkins, GitLab CI
Development Python, C++, Java

Use cases

Real-world applications

Documented outcomes from actual deployments.

1

Accelerated AI Unit Testing Process

AI automatically generates test cases from models and requirements, replacing the 20–40 minute manual process with a 5–7 minute AI-powered workflow at 85–90% accuracy.

Before

20–40 minutes manual test case creation per task, 1–2 month testing phases

After

5–7 minutes AI generation per task, 0.5–1 month testing phases

5–7 min vs 20–40 min manual
30–40% cycle reduction
2

MBD Check-Rule Automation

AI Check-Rule Model automates rule violation detection in Model-Based Design, achieving 70% coverage of rule checks and saving approximately 20 minutes per model.

Before

30–40 minutes manual rule checking per model, consistency issues, human fatigue errors

After

AI achieves 70% check coverage automatically; 20 minutes saved per model

70% AI rule coverage
35% effort reduction
3

AI Model Performance Validation

Comprehensive validation pipeline for automotive AI models — ensuring performance targets are met across accuracy, recall, and inference speed before production deployment.

Before

Inadequate baseline performance for real-time safety-critical requirements

After

5.99% accuracy gain, 49.15% recall improvement, 18.7× faster inference (quantized)

18.7× inference speedup
49.15% recall improvement

How we work

Implementation approach

1

Phase 1: Codebase Analysis

  • Analyze existing codebase and test patterns
  • Identify key components and testing requirements
  • Define test generation rules and constraints
2

Phase 2: Model Training & Customization

  • Train AI models on automotive cockpit code patterns
  • Customize for specific OEM requirements
  • Validate test quality and coverage
3

Phase 3: Integration & Validation

  • Integrate with CI/CD pipeline
  • Validate generated tests against existing test suites
  • Ensure compatibility with testing frameworks
4

Phase 4: Deployment & Optimization

  • Deploy to development teams
  • Monitor test generation quality and coverage
  • Continuously improve based on feedback

Explore more

Related services

Refactoring AI

Automated code review, MISRA-C compliance, and safe API porting for SDV migrations.

Data Analysis & ML

Predictive safety, driver context on Qualcomm, EV diagnostics, and drawing reviews powered by ML.

System Test AI

AI-assisted end-to-end vehicle validation and HIL/SIL simulation compliance.