AWS re:Invent 2024 recap for the Automotive Industry

At AWS re:Invent 2024, the AWS flagship annual conference held from December 2nd through the 6th, AWS unveiled the latest innovations and services with a week of keynotes, breakout sessions, product launches, and exciting demos. This recap highlights key announcements, customer stories, and demos presented that are relevant to the automotive industry, and organized across seven strategic workloads to help you quickly find what’s most relevant to your work.

Autonomous Mobility

Autonomous vehicle (AV) and advanced driver-assistance systems (ADAS) development requires a data-driven process that relies on hundreds of petabytes of drive data and complex tool chains. AWS CEO Matt Garman in his keynote announced the General Availability of AWS Trainium2 powered by Trn2 instances, the most powerful EC2 compute solutions for deep learning and generative AI training and inference.. With up to 4x faster speed, 4x more memory bandwidth, 3x higher memory capacity than predecessors, and up to 30% higher floating-point operations, these instances can deliver unprecedented compute power for ML training and generative AI and could support customers’ use of AI models to help them develop autonomous driving systems and vehicle navigation.

Continental in breakout session AUT307, shared how they are collaborating with AWS to accelerate their AV/ADAS development process in the cloud by offloading undifferentiated heavy lifting, such as streamlining data management and simplifying tool chain integration. The session also demonstrated how generative AI accelerates the process of identifying and managing the estimated 1–2% of relevant data within multi-modal datasets for model training, fine-tuning, and system validation. In session AUT318, Rivian shared how they accelerated their ADAS development on AWS and built an end-to-end pipeline for developing, simulating, and releasing new ADAS features built with support from AWS cloud services and partner solutions. Rivian also highlighted their approaches in various areas, from data ingestion to perception model training, simulation, and edge deployment. Rivian discussed the AWS services they used and shared insights on best practices employed to accelerate and scale new ADAS feature development, while optimizing data ingest, compute, storage, and simulation costs.

In session PRO202, Iveco Group showcased how they built a self-service development platform on AWS to enable faster development cycles and reduce hardware dependencies. The company’s Virtual Engineering Workbench fosters collaboration among developers, testers, and integrators by providing virtualized environments for digital cockpit development and automated testing. Additionally, the company’s Knowledge Management solution helps enhance productivity by streamlining access to technical information, and generative AI helps automate tasks like generating product data sheets from mechanical drawings. In AMZ201, Zoox, which recently deployed its purpose-built robotaxi in Las Vegas, shared how they built the ML Infrastructure that powers the autonomous driving of their robotaxis. This included how they collect data from robotaxis and how they designed their compute, training, and serving infrastructure to support autonomous driving and other ML use cases.

At the industry pavilion, the AWS automotive team showcased a demo on “Generative AI-Powered Development for AV/ADAS” which demonstrated the power of multi-modal, generative AI search in autonomous vehicle development. Customers challenged the system with unique prompts like “ducks crossing at an intersection” or “pedestrians with green hats,” and watched as the tool found relevant scenes from a vast dataset. Generative AI is revolutionizing how engineers train ML models for autonomous vehicles.

Software Defined Vehicles (SDV)

The domain of Software-Defined Vehicles (SDV) encompasses applications that prioritize transforming in-vehicle software to enhance connectivity, adaptability, and the integration of advanced technologies, driving innovation in next-generation mobility solutions. As cars become increasingly software-driven, with an expanding number of sensors and digital components, automakers must address challenges in software development, testing, and optimization. BMW faced the challenge of routing and resolving tickets during the testing and development phases of designing a new car platform. In session PRO201, BMW shared how over 140 software teams across various areas now use a generative AI large language model (LLM) that provides recommendations on the next best action, augmenting the previously manual process. This innovative solution streamlines BMW’s software-driven car development process.

In session AUT319, BMW Group also shared insights on the global software development platform within its automotive organization named the BMW Software Factory. After transitioning to AWS, BMW scaled their platform to run over 165,000 CI/CD builds daily, with up to 5,500 instances dynamically provisioned in parallel. Using strategies and fundamental architecture design choices that enabled a resilient platform, BMW optimized costs, improved developer productivity, and accelerated time to market for its more than 12,000 developers. Session ARC326 explored how chaos engineering enables organizations to proactively identify system resilience weaknesses, improve operational excellence, and enhance incident response capabilities. Through real-world scenarios executed with the AWS Fault Injection Service (FIS), the session highlighted BMW Group’s transformative journey, including key lessons on scaling chaos engineering, conducting large-scale chaos experiments in production, uncovering issues, and fostering a culture of resilience and continuous improvement.

In session OPN405, Toyota discussed how they elevated the developer experience using Backstage on AWS, an open-source framework that helps unify tools and services across cloud and on-premises infrastructure. By integrating Backstage into their platform engineering strategy, Toyota was able to streamline development workflows, fostering collaboration and improving developer productivity. The session also explored how generative AI models can be empowered through developer portals to unlock broader capabilities, showcasing how these strategies can transform the developer experience and enable innovation in software-defined vehicles.

The AWS demos showcased how to accelerate vehicle software development with generative AI, from using natural language queries to explore requirements to deploying AI-powered developer agents to write code and testing on virtualized hardware. These innovations reduce development cycles from weeks to minutes. The SDV journey of automotive customers has progressed from running virtual ECUs in the cloud to running co-simulations, system-level development and testing, and end-to-end pipelines for software development, CI/CD, and validation on AWS.

Connected Mobility

The rise of connected vehicles is transforming the automotive industry, enabling automakers to deliver seamless, data-driven experiences. In his keynote, Matt Garman highlighted how Generative AI could become a core component of every application. Amazon Bedrock simplifies building and scaling Generative AI applications by providing secure access to a wide range of models. But it is still surprisingly hard to find the right model for your use case with right expertise. Matt Garman announced several Amazon Bedrock enhancements such as Amazon Bedrock Model Distillation which transfers knowledge from a large, complex model to a smaller one. These distilled models are up to 500% faster and 75% cheaper, allowing automakers to deploy AI-powered services in vehicles with limited computational resources, such as infotainment systems and predictive maintenance modules.

In session AUT202, Honda demonstrated how they redefined the in-vehicle EV charging experience using Amazon Bedrock and AWS IoT Core. By integrating data from their connected vehicle platform, external sources, and generative AI, Honda delivers a personalized and intelligent charging solution tailored to the evolving needs of EV owners. At the heart of this solution, Amazon Bedrock powers personalized charging recommendations and route guidance by factoring in battery health, driving patterns, and available charging stations.

Session IOT321 showcased Toyota’s modernization of its connected vehicle platform using AWS IoT Core and MQTT 5. This transformation enables secure, scalable, bidirectional communication between vehicles and the cloud, delivering innovative applications. The session highlighted how advanced data collection using AWS IoT unlocks the full potential of connected vehicle data, supporting use cases such as remote vehicle controls, real-time data processing, fleet management insights, and EV battery optimization.

In the era of connected vehicles, leveraging real-time, data-driven insights is essential for enhancing customer experiences and improving operational efficiency. Session AUT311 explored Ford’s collaboration with AWS to develop the Event Store, a key component of Ford’s Transportation Mobility Cloud (TMC). This platform processes up to 4 TB of real-time data daily from over 20 million vehicles, providing insights on OTA updates, vehicle commands, and telemetry. By adopting AWS services, Ford built a petabyte-scale data lake using Apache Iceberg, improved data management and analysis, and reduced SLAs by up to 50%, meeting low-latency requirements.

At the expo, AWS showcased Virtual Driving Simulator featuring a vehicle chat assistant. Customers stepped into an automotive cockpit, drove through a virtual cityscape in the Carla simulator while real-time data from vehicle sensors was collected and processed using AWS IoT. After driving, they activated a voice assistant to access AI-powered insights about vehicle telematics and asked both general and vehicle manual-related questions. This immersive experience highlighted how AWS IoT and Generative AI are harnessing vehicle data to transform customer interactions.

Digital Customer Engagement

With the rise of on-demand content, automotive brands must better understand customer behavior to deliver personalized experiences tailored to individual lifestyles. By combining enterprise data with smart AI models, businesses can create engaging experiences that exceed customer expectations, foster loyalty, and drive continuous brand interaction. The true value of generative AI lies in integrating enterprise data with advanced models. Popular ways to add your data into a model is retrieval augmented generation (RAG) which can be done using Amazon Bedrock Knowledge bases, managing ingestion, retrieval and augmentation workflows. To improve model accuracy and prevent hallucinations, Amazon Bedrock Guardrails now supports Automated Reasoning checks, ensuring precise answers to queries about vehicle manuals, warranty policies, or repair procedures via voice assistants. Additionally, Amazon Bedrock now supports multi-agent collaboration, enabling complex workflows like scheduling vehicle service appointments while simultaneously providing cost estimates and nearby dealer options.

In his keynote, Amazon CEO Andy Jassy announced the launch of Amazon Nova foundation models, the new state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. These models allow users to generate visuals, videos, and fine-tune models across multiple domains. Amazon Nova can enhance generative AI tasks such as creating intuitive, engaging interactions between drivers and vehicles and dynamically personalizing the in-car experience based on user behavior and preferences.

In breakout session AUT205, CarGuru demonstrated how AWS machine learning services like Amazon SageMaker and streaming analytics improve the online car-shopping experience. Their platform powers vehicle recommendations, pricing insights, and an online marketplace. CarGuru helps car buyers with faster vehicle searches, well-timed promotional offers, and relevant vehicle suggestions.

At the expo hall, the Intelligent Call Center Assistants demo highlighted how Amazon Connect and Amazon Q enhance customer service in the automotive industry. The AI scanned internal documentation, responded to inquiries, and suggested real-time responses to agents. Customers tested this by asking questions and observing the AI-assisted replies. Lexus also showcased its GX550 vehicle with a digital concierge avatar chatbot powered by generative AI, answering queries and demonstrating the future of interactive in-car systems.

Manufacturing and Supply Chain

Advancements in data and AI are driving innovation in manufacturing and supply chain management. Amazon S3 Tables optimized for tabular data like sensor readings, stores data in Apache Iceberg format for seamless querying with popular query engines like Amazon Athena, Amazon EMR, and Apache Spark. Amazon SageMaker Unified Studio (preview) consolidates data engineering, analytics, and generative AI into one hub, providing analysts and data scientists seamless access to business data catalogs. Amazon SageMaker Lakehouse, new Apache Iceberg-compatible lakehouse, provides unified access to diverse data sources such as S3, Amazon Redshift, SaaS and federated data sources. Amazon SageMaker HyperPod accelerates model development by efficiently managing distributed training across clusters with hundreds or thousands of AI accelerators. AWS IoT SiteWise’s new generative-AI powered industrial assistant enables plant managers, quality engineers, and maintenance technicians to gain actionable insights, solve problems, and take decisions intuitively using natural language queries.

Quality remains critical in automotive manufacturing across all production stages. Volkswagen has transitioned from manual quality inspection methods to automated, machine learning–based methods, deploying use cases such as label inspection, assembly checks, and crack detection. In MFG207, Volkswagen shared how they scaled and sustainably implemented over 100 use cases powered by their Digital Production Platform (DPP). The session highlighted five years of DPP evolution and Volkswagen’s vision for smart manufacturing’s future.

In AIM117-S, Penske Transportation Solutions and Capgemini showcased their innovative connected fleet program, leveraging AWS ML/AI and generative AI services to revolutionize fleet management. The program enables predictive maintenance, cost savings, and automation efficiencies while processing terabytes of data at high speed. Penske and Capgemini also shared how their strategic adoption of cloud technologies introduced groundbreaking services, enhanced customer satisfaction, and unlocked new revenue streams.

In session AUT201, Toyota demonstrated how generative AI is driving innovation, enhancing productivity, and improving operational efficiency. They shared insights into AI-driven initiatives, including capturing institutional knowledge from retiring employees, reducing production line repair times, and decreasing battery waste. In session AIM236, Toyota presented its partnership with IBM to reimagine supply chain operations using AI on AWS. In session AIM383, Toyota Motor North America discussed a collaboration with Deloitte to enhance vehicle supply chain capabilities. This program streamlines production ordering, enables scenario simulations, reduces manual effort, and provides profit-optimized recommendations to boost vehicle sales.

In AIM120-S, Audi highlighted TenderToucan, an AI-powered tool that revolutionizes their tender process. By using LLMs to compare offers against requirements, TenderToucan reduces administrative workload, allowing employees to focus on analytical tasks. This innovation has significantly improved evaluation speed and accuracy, showcasing generative AI’s transformative potential in business processes and setting new standards for tender management in the automotive industry and beyond.

Sustainability and EV

Reducing energy consumption at operational and industrial sites is essential for organizations to achieve their sustainability goals. AI applications simplify the process, enabling organizations to optimize energy use more effectively. In session SUS304, Volkswagen Poznan shared how machine learning (ML) uses historical data from equipment controllers, combined with simulations and forecasting, to drive sustained energy efficiencies across operations and facilities.

In session BIZ219, BMW highlighted their collaboration with AWS to advance sustainable and transparent supply chain management in the automotive industry. They discussed the need for a comprehensive product carbon footprint (PCF) solution that includes supplier emissions data collection, augmentation, aggregation, validation, and auditing to streamline the exchange of supplier certificates.

At the expo, visitors explored the Intelligent EV Charging System demo, featuring a smart, scalable, and serverless EV charger fleet management system. The demo showcased a user-friendly interface for simplified charger management and demonstrated how Generative AI supports configuration, diagnostics, and troubleshooting. An intelligent travel advisor also generated optimal charging plans and suggested value-added services for EV owners. Customers experienced these innovations firsthand, highlighting AWS’s role in shaping the future of automotive technology.

Migration and Modernization

AWS CEO Matt Garman announced Amazon Q Developer Transform for Mainframe which can transform mainframe applications, reducing time spent on undifferentiated tasks and enabling developers to spend more time on value-added activities. It can help transform IBM z/OS mainframe applications through code analysis and migration planning. Even for often poorly documented legacy systems, Q can build real-time documentation for millions of lines of code of legacy systems and reduce multi-year effort modernization projects into multi-quarter efforts, cutting timelines by more than 50%. Amazon Q integrates seamlessly with the AWS Console, Slack and all the popular IDEs like Visual Studio, VS Code, IntelliJ, while offering deep integration with GitLab. As maintaining and modernizing applications consumes significant resources, Amazon Q Developer Transform can now modernize .NET applications from Windows to Linux in a fraction of the time, up to 4x faster. Amazon Q Developer Transform for VMware workloads can transform VMware workloads to cloud native architecture by identifying application dependencies and generating a migration plan.

In session MAM238, Toyota shared exciting progress in its supply chain transformation. After 45 years of running its supply chain on mainframes, Toyota faced challenges migrating off legacy systems, including outdated documentation and the impracticality of reverse engineering hundreds of millions of lines of COBOL code. By collaborating with AWS, Toyota enhanced Amazon Q to evaluate COBOL code and generate comprehensive business and technical documentation. Amazon Q’s document generation agents processed 400 COBOL programs (375,000 LOC) in just 5 hours with up to 90% accuracy, allowing Toyota to document supply chain mainframe programs in days rather than months. Toyota’s innovative use of generative AI to accelerate mainframe modernization has the potential to cut migration timelines by up to 50% while introducing new business capabilities.

Conclusion

Customer stories from industry leaders like BMW, Toyota, Volkswagen, Honda and Rivian, along with exciting demos and keynotes, showcased how AWS drives innovation across the automotive industry. From enabling autonomous driving and accelerating software development to modernizing supply chains and fostering sustainability, AWS is at the forefront of the automotive industry’s transformation.

Now is the time to explore how AWS can help accelerate your automotive innovations. Visit AWS for Automotive to learn more, access resources, and start your journey toward the future of mobility. Let’s drive the automotive industry forward together!