Building AI applications for Physical Intelligence using AWS with ADI

ADI combines analog, digital, AI, and software technologies into solutions that help drive advancements in automation, robotics, mobility, healthcare, energy, and data centers. To accelerate innovation, the company was inspired to build a developer platform that would provide its teams with the tools they need to rapidly create AI applications. At the time, AI development and deployment were decentralized across teams, prompting ADI to pursue a more standardized approach that could support scalability and consistent governance across the organization. The company wanted to remove the technical hurdles that were slowing AI development while also helping field teams build trust with customers in AI solutions uniquely aligned to product performance specifications.

While designing a new platform, ADI drew from its previous experience on AWS. “In terms of rapid prototyping, there’s no comparison to serverless,” says Shane Clifford, senior director of software tools and platforms at ADI. “We gave ourselves the time to figure out serverless, cloud-based, fully managed services, and the AWS Shared Responsibility Model. Since then, our mindset has changed from traditional stacks, and we’ve never looked back.”

ADI prioritized the developer experience by architecting its self-service platform—AI Studio—on open source standards, including PyTorch and TensorFlow frameworks and Hugging Face datasets. “Our cloud development teams are passionate about the developer experience and getting out of our own way with automation,” says Clifford. “We want to bring that mentality to every consumer, every product, and every tool that we develop.”

The company builds its services based on an API-first and event-driven architecture. As such, ADI chose to automatically scale its platform using AWS Fargate, a serverless compute service for containers, and AWS Lambda, empowering its teams to run code without thinking about servers or clusters. To create a fully automated processing pipeline, ADI adopted AWS Step Functions for serverless orchestration for modern applications.

ADI integrated AI Studio directly with its source code management environment, giving its developers seamless access to a portfolio of enterprise-ready foundation models. These models are optimized for edge devices and can be used across various use cases, including anomaly detection, denoising, and audio classification.

The company’s software and operations teams also collaborated to create a repository of templates that its teams can use to quickly provision resources. To maintain governance over its environments, ADI uses AWS Identity and Access Management (AWS IAM), which helps the company securely manage identities and access to its AWS services and resources.

In less than 4 hours, ADI created a proof of concept for AI Studio. Within 6 months, its teams worked together to finalize the platform and move to production. “We didn’t have to over-engineer things,” says Maria Butler, senior manager for AI/ML software at ADI. “We were able to start with what we needed and then evolve the design over time. We could move quickly to deliver something to our users.”

Deploying AI Studio has helped ADI explore new edge use cases while reducing its total cost of ownership by an estimated 67 percent. Instead of procuring hardware and relying on enterprise storage, ADI’s developers can spin up an environment to test their ideas in 20 minutes or less. The platform offers secure access to a range of models optimized for time-series, audio, and sensor data, accelerating the development of real-world edge applications.

ADI has started using the platform to prototype AI applications with its customers for a variety of emerging use cases, including autonomous technology, robotics, and more. In the past, field teams could only demonstrate ADI’s capabilities through slideshow presentations. This advancement further improves ADI’s ability to showcase tangible value from the get-go. “We achieved one of our early adoption goals to empower our customers and inspire them to build AI-driven solutions that can take their products and their companies forward with us,” says Yuval Zukerman, director of edge AI partnerships at ADI.

Moving forward, ADI will continue adding capabilities to AI Studio. “Using AWS fully managed services helped us maximize the impact of our team. AWS gave us the tools we needed to quickly build the right solution, the right way, while letting our data scientists focus on what they do best,” says Clifford.