A Year After Acquisition, Renesas is Redefining AI Solutions Portfolio with Reality AI Software. It’s been just more than a year since Renesas announced our intent to acquire Reality AI. We celebrated that milestone at this month’s Sensors Converge conference in Santa Clara, Calif., where we showcased Reality AI’s artificial intelligence (AI) and tiny machine learning (TinyML) solutions incorporated across multiple applications. Reality AI is one of the critical component of Renesas’s overarching AI strategy. You will hear more from us over the next several months as we host a panel during July’s SEMICON West/DAC conference in San Francisco, where we will join industry colleagues to examine how AI is changing the face of semiconductor design and manufacturing. That will be augmented by a series of AI-themed podcasts and webinars that will culminate in the fourth quarter with Renesas AI Live.
We sat down with Mohammed Umar Dogar, vice president of global business development and ecosystem at Renesas. Mohammed shared his thoughts on how the Reality AI acquisition has positioned our company to integrate AI across new tools and system-level solutions to improve the development journey and make our customers lives easier.
Please remind us, what was it about Reality AI that appealed to Renesas? What did they bring that wasn't available to Renesas before the acquisition?
Mohammed: The relationship with Reality AI dates back to 2021 when we were ramping up the technology ecosystem for our RA microcontroller product line. We were looking at system-level enablers to help customers make something useful with highly efficient edge and endpoint computing.
Reality AI stood out because of their unique ability to combine signal processing and other proprietary technologies with ML to build super-efficient machine learning models that were highly effective and extremely small.
Vision and voice applications get a lot of press, but there is potential with a great many non-visual/non-verbal ML applications in the industrial and automotive sectors. We knew that our customers could benefit from TinyML models capable of anomaly detection, classification and regression based on the kind of data that come from real-time, high frequency sensors typically used in these applications.
What set Reality AI apart from the other ecosystem partners you vetted?
Mohammed: Out of the seven or eight companies we looked at, Reality AI had the best combination in terms of skill set, team capabilities and critical core AI technology. And more interestingly, they had successful engagements with leading industrial and automotive customers around the world, which was good validation of their technology. It was also an indication that they had implemented the kind of quality control that industry requires of its vendors. When we looked at them closely, we saw in-depth experience in several vertical markets — that is usually very difficult for startup companies to achieve.
How do things look a year after the acquisition? What have you learned?
Mohammed: The integration is going incredibly well. Reality AI is now exclusively available across Renesas’ broad range of MCUs and MPUs and supports everything from our 16-bit RL78 microcontrollers all the way to our 64-bit RZ microprocessors and the RX and RA families in between.
Also, when we announced the acquisition, we said we would be creating an AI center of excellence at Reality AI’s former headquarters in Columbia, Md. A year later, the team is developing vertically-layered software tools and an ecosystem for customers to build AI and machine learning solutions across our hardware portfolio. In this short time, they have already focused on several aspects of improving the overall customer experience when it comes to designing for AI and machine learning.
Why is that significant?
Mohammed: Traditionally, the AI tool chain and the embedded tool chain are very disconnected. One of the things we worked hard to understand was how to marry these two domains so that Reality AI Tools® worked seamlessly with the Renesas e² studio, which is our Eclipse-based integrated MCU development environment.
Very early on, we were able to build an API-based context exchange between these two tools, and now we have a complete road map. Obviously, there's a lot of work remaining to enable a completely seamless experience, but we’ve certainly tied in the two tools from a contextual exchange perspective so that customers can easily transfer projects between the two. That's a good example of what we've done to ease the developer’s journey.
What other Reality AI achievements can you point to over the past year?
Mohammed: One is that we built a portfolio of solutions at different levels, including simple application examples that demonstrate how a particular AI/ML solution would apply to a particular use case – for example, detecting an unbalanced load condition on a motor that shares electrical data with the motor controller. That application example serves to illustrate what customers can do with a new toolbox add-on to Reality AI software that integrates with Renesas Motor Control Kits and allows Reality AI models to be integrated with the motor control process (RealityCheck™ MOTOR). This enables a sensor-less ML model that can detect conditions without any addition to the bill-of-materials, and enables customers to add predictive maintenance or condition monitoring functionality with nothing but a firmware update.
We've also created the Holy Grail of vertical integration with what we call RealityCheck™ HVAC, an endpoint AI-enabled solution for creating smart, self-diagnosing air conditioners, heat pumps and refrigeration systems. This is a completely integrated solution that includes hardware, software, reference design, a selection of sensors and a production-grade data set. We instrument customers’ HVAC systems in a professional lab and collect high quality data to build and calibrate product-grade models.
So, to recap, our Reality AI solutions include application examples, which are mainly for demonstration purposes. Then there are toolboxes that help customers working in specific domains with data pipelining and collection. And then we have completely integrated solution suites.
What has been the biggest surprise for Renesas as it assimilates Reality AI into your business lines?
Mohammed: The biggest surprise has been the reaction from our customers. We expected there to be interest in Reality AI following the acquisition, but we’ve been blown away by scale of that interest. We now have customers working with Reality AI software and solutions in the U.S., Europe, Japan, China, Taiwan, even Africa. We have augmented their team with resources around the world to meet that demand and provide needed support. We are winning competitive sockets by combining Reality AI offerings with Renesas’ best-in-class MCUs and MPUs, and the first end-customer products featuring Reality AI firmware running on Renesas processors will be coming to market soon. We’re just getting started!