TCS Enhances Automotive Software Testing with NVIDIA AI, Doubling Speed
Jessie A Ellis Nov 22, 2024 19:15
Tata Consultancy Services (TCS) leverages NVIDIA's AI technology to double the speed of automotive software testing, enhancing customer experience and accelerating the software engineering lifecycle.
Tata Consultancy Services (TCS) has made significant strides in the automotive software industry by utilizing NVIDIA's generative AI technology to double testing speeds, according to a recent report by NVIDIA. This advancement is part of TCS's broader strategy to enhance customer experience and streamline the software engineering lifecycle in the automotive sector.
Transforming Automotive Software Development
The automotive industry is undergoing a transformation from mechanical to software-driven systems, and generative AI plays a crucial role in this shift. By employing AI-based algorithms, TCS aims to improve decision-making processes, personalize user experiences, and develop fully autonomous vehicles. This technology allows for the generation of comprehensive datasets that enhance vehicle personalization, from in-car personal assistants to entertainment recommendations.
Accelerating Software Engineering
As vehicles become increasingly complex with millions of lines of code, the demand for rapid development and deployment of automotive features has surged. TCS addresses this challenge by applying generative AI to accelerate various stages of the software engineering lifecycle, including requirement analysis, design, development, and validation.
TCS's Automotive Gen-AI Suite
To achieve these goals, TCS has developed the Automotive Gen-AI Suite, which incorporates TCS-patented algorithms and NVIDIA technologies. This suite employs large language models (LLMs) trained with the NVIDIA NeMo framework, fine-tuned using domain-specific datasets. The suite is designed to generate unit-level test cases, significantly reducing the time and cost associated with manual test case creation.
Optimizing Test Case Generation
Generating test cases from unstructured requirements is a labor-intensive process. However, TCS's integration of NVIDIA technologies automates this task, allowing LLMs to produce test cases with minimal manual intervention. The models are fine-tuned using techniques like Low-Rank Adaptation (LoRA) to optimize performance and accuracy. TCS's deployment on NVIDIA DGX H100 systems ensures low latency and high throughput, enhancing the overall efficiency of the test case generation pipeline.
Benchmarking and Model Selection
TCS conducted comprehensive benchmarking to select the best models for test case generation, focusing on accuracy, decision coverage, and Modified Condition Decision Coverage (MCDC). The Llama 3 8B Instruct model, fine-tuned with NVIDIA NIM, showed superior performance, achieving higher accuracy and coverage compared to pretrained models.
Future Prospects
With a focus on generative AI, TCS plans to advance its capabilities by exploring conversational LLMs, visual LLMs for context understanding, and image-based models. The company is also set to utilize NVIDIA Blueprints to further refine the software engineering lifecycle, promising continued innovation in the automotive software industry.
For more detailed insights, visit the full article on NVIDIA.
Image source: Shutterstock.jpg)