Woven by Toyota saves millions and scales autonomous driving with Union.ai
Challenge
Autonomous driving AI requires highly scalable infrastructure.
Woven by Toyota (WbyT), part of the Toyota Group, builds advanced automated driving (AD/ADAS) technologies and Arene OS, its automotive software development platform. These workloads are among the most demanding in the industry, requiring real-time decision-making, petabyte-scale data processing, GPU-intensive training, and strict safety and regulatory compliance.
The team initially used Flyte for orchestration, but as their autonomous driving programs expanded, maintaining internal infrastructure became a bottleneck. They needed more than workflow orchestration—they needed enterprise-grade AI development infrastructure capable of handling massive parallelism, data throughput, GPU scaling, reproducibility, and deep integration with internal systems on AWS.
They sought a managed platform that could support their long-term autonomous vehicle roadmap without requiring continuous investment in platform maintenance.
“Union.ai support has proven to be critical for us at the time that we needed to significantly scale up our data processing needs to meet major milestones.”
Alborz Alavian
Senior Engineering Manager, Woven by Toyota
Solution
A unified platform for scalable AI development.
In 2023, Woven by Toyota migrated to Union.ai, a fully managed, production-grade platform designed for advanced AI workloads. Union became the foundation for WbyT’s AI development infrastructure—powering everything from data annotation to perception labeling, dynamic scene processing, and GPU-accelerated model training.
Union.ai now manages petabytes of sensor data, hundreds of thousands of node hours, and large-scale distributed workloads across WbyT’s autonomous driving programs. It abstracts the complexity of cluster management, scheduling, GPU allocation, and caching, allowing ML engineers to focus entirely on developing safer, smarter driving systems—not infrastructure maintenance.
Union’s managed infrastructure also ensures reproducibility across historical sensor data, unlocking continuous improvement of AV models through large-scale backfilling and recomputation.
By offloading Flyte maintenance and associated DevOps overhead, WbyT gained a reliable, scalable AI platform aligned with its long-term velocity and safety goals.
faster ML iteration cycles
saved annually via optimized compute and scheduling
of parallel workers powering large-scale ADAS pipelines
Results
Union.ai accelerated innovation and reduced AI development costs.
Union.ai transformed the speed and economics of WbyT’s autonomous driving programs. High-throughput data processing and GPU utilization accelerated model iteration by more than 20×, enabling faster deployment of ADAS features and perception models.
Advanced scheduling, caching, and spot instance optimization drove millions in annual cost savings. With thousands of parallel workers and massively scalable backfilling, WbyT can now continuously retrain and refine models using historical sensor data.
By removing Flyte maintenance from WbyT’s engineers, Union.ai dramatically reduced DevOps overhead and allowed teams to focus on innovation, safety, and model performance rather than infrastructure management.
As WbyT continues to invest in the future of mobility, Union.ai provides its foundation for AY development infrastructure, where data processing, model training, real-time inference, and infrastructure scale come together seamlessly.
