Nvidia Snatches Safety Guru from Intel

Article By : Junko Yoshida

With fierce competition to add autonomy to everything, safety is critical. Companies are waging a talent war over developers with expertise in system safety.

At a time when companies in the electronics industry are competing fiercely to add “autonomy” to everything from end-point devices to robotics and autonomous vehicles, they are also waging a talent war over data scientists and AI algorithm developers.

Emerging next is the recruitment battle over safety experts.

Nvidia this week snatched up Riccardo Mariani, formerly chief functional safety technologist at Intel Corp., and an Intel Fellow, to be its vice president of industry safety. The position is a new role for Nvidia, added to augment its “world-class safety team,” according to Nvidia. The goal is to institute “an overarching safety view across multiple domains and technologies.”

Mariani, a guru in functional safety and integrated circuit reliability, is widely recognized for his leadership role in the development of a variety of industry standards including ISO 26262. He is expected to boost the credibility of Nvidia’s commitment to the reliability and safety that must come hand-in-hand with autonomy.

Why Nvidia?

EE Times interviewed Mariani by email. He told us, “Among the multiple offers received, I selected Nvidia for its strong commitment on safety and recognized presence in safety critical automotive and embedded markets.”

He noted “the well-defined and established safety process and culture” already established at Nvidia. Asked about his mandate, he said, “I will drive safety alignment across business units (Auto, Robotics, IoT, etc.), developing cohesive safety strategies and cross segment safety processes, architecture, and products allowing work to be leveraged across our hardware and software platforms.”

He added, “I will also support the existing team in evangelizing key safety elements, both internally and externally, with customers, partners, regulatory agencies, government, etc.”

Mariani joined Intel when Intel acquried Yogitech in 2016. Mariani was co-founder and CTO at Yogitech, which was then described as a leading provider of functional safety technologies. He invented the company’s flagship faultRobust technology and related products.

Further, Mariani led Yogitech to pioneer the certification of semiconductor intellectual property for the highest level of safety integrity, and to introduce electronic design automation tools specific to functional safety.

Asked how his role at Nvidia will differ, Mriani told us, “I want to continue my focus on the overall ‘end-to-end dependability’ aspect of intelligent systems.”

He explained that functional safety must be combined with “reliability, availability, maintainability, security and time sensitivity. All these properties must coexist to provide trustable, deterministic and safe services to the community.”

Recommended
Improving Safety Standards for AV

Asked about his activities in various industry standards committees, he stressed that he has no intention to slow down. He noted, “I will certainly continue my active role in various industry standards committees and join the other Nvidia team members in representing the company in ISO 26262, ISO 21448 (SOTIF), IEC 61508 and others.”

Mariani added, “I will also continue my engagement in IEEE Computer Society as VP for standardization activities and chair of IEEE CS Special Technical Community on Reliable, Safe, Secure and Time Deterministic Intelligent Systems.”

Mariani, who was very active in putting together ISO 26262 functional safety standards, is also involved in developing the emerging standard called ISO 21448 — Safety of the Intended Functionality (SOTIF). In a previous interview with EE Times he said, “We don’t think ISO 26262 is enough” to guarantee safety. Noting that the industry has seen the “Uber accident and other events in which autonomous-driving technologies were misbehaving,” Mariani made the case for why autonomous vehicles (AV) must go beyond ISO 26262.

SOTIF, he noted, is necessary to avoid unreasonable risks for ADAS and AVs — even in the absence of malfunctions by hardware and software — that might pose trouble on the road. ISO 21448 is said to complement ISO 26262, picking up where ISO 26262 leaves off.

Nvidia has won the favor of the financial community with its GPU technology, viewed effective in various artificial intelligence applications — especially in training, but also in self-driving vehicles.

It's hard, however, not to overstate the importance of safety of AVs, enabled by the AV full stack developed by AV tech companies.

The Uber self-driving car involved in the first fatal accident during robocar testing did not use Nvidia’s autonomous vehicle computing platform (instead, Uber used Nvidia’s standard GPU). Nevertheless, Nvidia suspended self-driving tests across the world a week after Uber’s vehicle killed a woman crossing a street in Arizona in March 2018.

Nvidia isn’t standing still, though. Last month, the GPU giant announced a new partnership with Volvo to develop a next-generation decision-making engine for fully autonomous commercial trucks and industrial service vehicles. The partnership will use Nvidia’s Drive artificial intelligence platform, which encompasses processing data from sensors, perception systems, localization, mapping and path prediction and planning.