The car keeps learning after delivery.

A car that is finished on the day it is sold

Sonatus starts from a complaint that automotive engineers know well. Most vehicles still log data the way they were programmed at the factory: engineers choose months ahead which signals a car will record, and changing that means rewriting software and pushing an update. Alexandre Corjon, who runs engineering at Sonatus, added in July 2026 that validation still leans on physical fleets and long test cycles, even for minor software changes. The company's June 2026 launch post put the cost in human terms: engineers spend significant time gathering data, investigating issues and deploying fixes after problems occur.

It’s like a laptop

Jeff Chou · CEO and co-founder, Sonatus · Interview with The Drive · 2022-11-16

The belief behind the company is that a car should behave less like a finished product and more like a computer that improves through its life. Co-founder and CEO Jeff Chou has used one image for five years: the 2021 funding release spoke of turning vehicles into "data centers on wheels", and at CES in January 2026 he called the car "a moving data center on wheels". His co-founder, CTO Yu Fang, describes the company setting out from its first day to carry data-center ideas into the car: software-defined networking, storage and compute.

Software first, then the brain

What changed in 2025 and 2026 is the word the company leads with. Its 2024 releases spoke of software-defined vehicles (SDV), cars whose functions are controlled and updated by software. Its 2026 company page opens with "cloud-to-edge AI for smart vehicles". Yu Fang explains the order: without the SDV layer's ability to collect, govern and execute on data, AI cannot be deployed in the vehicle, so the plumbing has to come first.

SDV is the infrastructure; AI is the brain

Yu Fang · Co-founder and CTO, Sonatus · Interview with Gasgoo · 2025-12-18

Why now, in the company's telling, comes from the car's hardware. Suppliers such as Schaeffler now sell central control units that help automakers centralise their software, and Sonatus argues that much useful AI work can run on the compute a car already carries, without waiting for new chips. Generative AI also lets an engineer ask a car for data in plain language. A forecast that Sonatus sponsored puts the in-vehicle AI market in Europe, North America and Japan at $43 billion in 2025 and $238 billion in 2030; it is a forecast of a market, not of Sonatus revenue.

One fault, followed through the loop

Take an illustrative case: an automaker's engineer sees the same fault across part of a fleet. To observe, a Collector policy decides which signals to gather from the affected cars and when, so nobody has to log every byte. To analyze, Fastlane Insight compares behaviour across vehicles in the cloud and helps form a hypothesis to check on a test car. To act, Fastlane Edge puts a model, a virtual sensor or new diagnostic logic into the vehicle, and an integration engineer fits it to the car's compute, its network and the customer's requirements. Finally the change is proven on benches and test vehicles, and what was learned shapes the next policy.

Human judgement stays at every step. In the Nissan proof of concept, a senior Nissan research executive called AI a tool for engineers, not a replacement, and the engineer leading the project said her colleagues still make the decisions, only faster and with more reliable data. Yu Fang sets the bar for diagnostics: there, AI must guarantee deterministic correctness.

For someone joining in Korea, the act and prove steps are the job. Five of the seven open Korea roles sit in the company's Customer Success department and make Sonatus modules run, pass tests and pass audits inside a specific automaker's architecture; the other two run the build servers and lab hardware those engineers depend on.

Four products in one loop

Sonatus sells the Fastlane Platform, named in June 2026, as four products that work alone or together: Collector gathers vehicle data, Insight analyses it in the cloud, Edge acts inside the vehicle, and Copilot packages Collector and Edge for test cars. Buyers are automakers (OEMs) and their system suppliers (Tier-1s). Older articles use older names: Insight grew out of AI Technician, and Edge out of AI Director.

Fastlane Collector decides what a car records. Engineers write targeted collection policies, with generative AI helping to turn plain language into a policy; the policy is deployed into the vehicle and runs entirely there. The product page claims compression of up to 28 times. The company says Collector has run in production since 2020 and, as of June 2026, supports more than 8 million vehicles across multiple automaker programs.

Fastlane Insight is the analysis layer. It correlates vehicle data, logs, diagnostics and engineering knowledge to reach a root cause faster, and keeps diagnostic knowledge so it can be reused across programs and vehicle generations instead of living in a few engineers' heads.

Fastlane Edge is the execution layer inside the car. It deploys AI models, virtual sensors and diagnostic logic into production vehicles, manages rollouts over the air and supports several kinds of ECU, the small computers that run a car's functions. Its key claim is that it uses the compute a car already has; partners' models run on it, and Michelin deployed its tire models through the earlier AI Director. Fastlane Copilot is a gateway box on an Arm-based automotive processor, pre-loaded with Collector and Edge and wired to the car's CAN-FD, Ethernet and LIN networks, so an automaker can use the platform on prototype cars without embedding Sonatus software.

The anchor customer

The deepest customer relationship is with Hyundai Motor Group. A partnership across Genesis, Hyundai and Kia was announced in August 2021, starting with the Genesis GV70, and in November 2021 the Genesis GV60 became the first production vehicle with the second generation of the Sonatus platform, then called Digital Dynamics. The work produced a central communication unit (CCU) that carries over-the-air updates, data collection and remote diagnosis; Hyundai Motor Group's own newsroom names Sonatus as its partner on it, and it won a 2022 PACE Innovation Partnership Award.

It is quite meaningful to win such an award at a time when we are striving to transition to SDVs

Hyung Ki Ahn · Vice President, Electronics Development Group, Hyundai Motor Group · Hyundai Motor Group newsroom · 2022-09-21

The relationship reaches beyond passenger cars. A company case study describes Hyundai Commercial Vehicles launching the XCIENT Fuel Cell Truck and ELEC CITY Bus against traditional development cycles of 24 to 36 months. Hyundai adopted Sonatus Foundation for a modular vehicle network shared across vehicle lines and put Collector AI into its Blue Link Truck & Bus service. The case describes potential savings during design, alongside trigger-based diagnostics, predictive maintenance and driver logging. Those are design-phase estimates and described service capabilities, rather than measured benefits for every vehicle or fleet.

What was measured, and how far it goes

The clearest before and after comes from Nissan Technical Centre Europe, which announced a development partnership in December 2025. Before, data loggers on prototype cars recorded from the moment the battery was connected, and engineers saw the data only after the car reached the end of the line and someone uploaded the logger. With Collector AI and AI Technician, policies are deployed remotely and test data is analysed with AI.

The Nissan development-centre engineer described the result as a proof of concept: a root-cause investigation moved from two weeks to two days. The December 2025 announcement likewise calls these early trials of a tool still under development. In the April 2026 customer transcript, remote access and shared data let colleagues in different locations review the same evidence while engineers retain responsibility for decisions.

Signet reads these examples at their stated scope: a production component recognised with an industry partnership award, a commercial-vehicle architecture with estimated design savings, and a Nissan proof of concept whose investigation-time result recurs in company materials. For an applicant, the useful question is how the team reproduces an outcome across a new vehicle program. The Korea integration and compliance roles supply part of that answer through automated tests, documented interfaces, static analysis and release evidence.