Sonatus Launches AI Director to Integrate More AI in More Vehicles

Debuting at IAA 2025, the software platform acts like a middle manager, allowing AI services to find a home without adding more hardware.

September 3, 2025

Sebastian Blanco

Sonatus AI Director takes advantage of underutilized compute capabilities throughout a connected vehicle to bring more AI to more vehicles. (Sonatus)

Vehicle software company Sonatus wants the automotive industry to rethink the role of the middle manager. At least, when those managers are AI-based. The company unveiled its new AI Director end-to-end tool-chain software product in early September, calling it a “game-changing platform” that will allow OEMs to deploy AI at the vehicle edge without adding more compute. Sonatus will demonstrate AI Director for the first time in public at the IAA Mobility conference in Munichnext week.

Sonatus AI Director goes through five stages - collect and curate data, select and train/import a model, optimize and validate, deploy and integrate and then monitor and evaluate - before iterating and cycling through the steps again. (Sonatus)

Sonatus’ basic message with AI Director is that AI has already been in our vehicles, but there’s room for growth outside of the current focus on powertrain and driving operations, space for smaller suppliers to contribute to the automotive AI ecosystem.

“There’s a whole class of AI that is not related to driving or assisted driving, for example, component or vehicle optimization,” Jeff Chou, Sonatus CEO and co-founder, told SAE Media. “Think about how you would potentially tune or manage or predict your battery range better. What about tire optimization? How would you predict tire wear better? How would you tune your cameras or compensate these cheaper, consumer-grade, commodity cameras in the wake of varying temperature or weather?”

Another area where Sonatus believes AI can be useful is in anomaly detection, personalization or for dealer-owner coordination or cybersecurity. “The footprint of vulnerabilities is increasing,” Chou said. “It’s a natural use case for AI.” Earlier this year, Sonatus introduced its AI Technician product that helps consumers and engineers diagnose problems in a vehicle.

Partnering up

AI Director comes integrated with NXP’s eIQ Auto ML software. NXP is one of Sonatus ’launch partners, alongside VicOne, Compredict, and Qnovo. Qnovo, for example, tested AI Director with its Health & Safety Diagnostics (HSD) software to shorten the time-to-market for its battery safety software by identifying potential failures weeks in advance. Meanwhile, Compredict used AI Director to develop an AI-based Virtual Sensor that replaces standard ride height sensors with a software-only solution that will allow OEMs to add automatic headlight leveling to passenger vehicles (mandated in 2027 in Europe by UN R48-09) while saving money - a claimed up to $20 per vehicle - by lowering bill-of-materials (BOM) costs.

“There’s a whole class of smaller, more narrowly focused AI applications that offer a lot of value across the board, not just to the driver, but the entire ecosystem, the entire value chain,” Chou said. “It’s not just your OEMs doing autonomous vehicles, and it’s not just Tier One suppliers doing ADAS. You have a whole new set that is broader and wider that can participate and innovate.”

To facilitate that innovation, Sonatus said it wants AI Director to give OEMs and suppliers “an end-to-end toolchain for model training, validation, optimization, and deployment,” both by bridging silos within an OEM as well as gaps between a model provider and the OEM itself. The low-requirement software is silicon- and hardware-agnostic.

There are two pieces of AI Director: an already-existing part in the cloud that processes data, including model training, optimization, and monitoring. The Sonatus AI Director supports physics- and neural network-based models and Small and Large Language Models (SLMs/LLMs). The piece in the car puts the model into what Chou called the “connective tissue” of all the signals moving across all the actuators and thus providing an environment for the lightweight, in-vehicle model to operate. Think of the cloud part as a management console, John Heinlein, Sonatus CMO, told SAE Media. “Everything we’re talking about in this announcement is about the AI that runs in the vehicle. If you think about software-defined vehicles, they are maybe not realizing their full complexity. They oftentimes already have the ability to do isolated workloads, just like your phone is not running one application, it’s running 20 applications at the same time and they don’t know the other ones are there. What we’re saying is, if you have spare headroom in the existing systems, we’re giving you a fresh, new way to consume it with new types of computes. If you have a little headroom, we’ll run a little. If you have a lot, we’ll run more.”

As an example of how an AI cybersecurity model operates under AI Director is that the software tells the model to run a few times a day to check for new threats, perhaps when the car is idle. “The reality is that AI is a new way of coding algorithms and a new way of solving problems that’s fundamentally different,” Heinlein said. “There’s a spectrum of different AI, from more rudimentary models all the way through LLMs. Our goal here is to provide a framework to allow all of those different kinds of models to be deployed more easily, more consistently and in a more cost-effective way.”