Palette Edgematic: Create, build, and prototype computer vision ML pipelines in minutes vs months

Krishna Rangasayee was created to solve any computer vision application challenge with 10x better performance per watt, delivered in a push-button experience. Today, we’re advancing our vision for a push-button approach to ML at the edge by unveiling Palette Edgematic,’s free no code “drag and drop” software offering designed for any organization to get started creating Edge ML designs in minutes.

The idea for this product has long been in my mind since starting

Early on, we knew that what worked in the cloud was never going to work for edge AI and ML. To enable customers at the edge, we purpose built a system – both hardware and software from the ground up, in parallel – that could exceed the performance of the cloud and adhere to the power constraints of the edge. Ultimately the software running on this system, today known as Palette, powered by the SiMa MLSoC, would become the simplest and easiest software experience ever created for the embedded edge. It would intuitively understand that frames per second may have worked for the cloud, but at the edge, the defacto standard would soon become frames/second/watt (FPS/W).

We have come a long way since then and are proud to credibly say that today, is leading the way in setting a new performance and efficiency standard at the edge. SiMa regularly outperforms our primary legacy competitor in customer trials on a FPS/W basis…

And, twice now when we’ve gone head-to-head versus our primary competition Nvidia (OrinX and AGX) as measured via third party validated benchmarks facilitated by the MLCommons organization both in April and again in August of this year, we have won.

SiMa’s edge ML leadership doesn’t stop there. We have always said SiMa is a software company building its own silicon. And aside from FPS/W dominance, our second area of laser focus has been to make the entire ML edge experience simple and seamless at the push of a button. We believe in enabling a pushbutton experience, making ML effortless for all, and market our Palette software as being ML Software ‘done right.’ With Palette Edgematic we advance our software vision even further.

One of our earliest observations (and validated by extensive end user conversations) was that software will be the real differentiator in the race to win the edge ML market. As I’ve said before, Silicon Valley has been living up to its name too much with the belief that every problem is solved by improving the semiconductor chip design. However, if you talk to developers and understand their pain points, they visualize their problem through the lens of software.

This was the genesis for Palette Edgematic. If you’re familiar with the phrase “look ma’, no hands,” – one of the first visuals I made when founding the company was how to create SiMa’s own version of “look ML, no code.” It hadn’t been done in decades by semi or any chip companies, who traditionally were known for their lackluster software – let alone with a focus on the edge.

Together, SiMa brings a radically simpler way to program edge ML combined with best in class FPS/W execution, all made possible by a purpose-built technology architecture that will:

  • Scale with any amount of complexity;
  • Give machines and devices the speed, efficiency, and reasoning of real-world
    environments (in many cases, in a small physical form factor typically necessary at the
  • Give organizations the technical expertise, resources and talent to pursue their AI
    projects. While there’s more than 30 million developers globally, there are far fewer
    skilled in AI development and the learning curve to AI can be costly with a majority of
    talent locked up in today’s hyperscalers;
  • Provide any organization with an onramp to approach ML and the means to AI and
    ML-enable their operations, products and services in new and even superhuman ways.

Again, in minutes. Not weeks or months. Production-ready ML applications built in minutes with Edgematic. So any individual or organization who has been held back by the complexity of executing ML at the edge, can get started, scale quickly and develop applications that can change the world.

Here’s a glimpse of Edgematic at work:

Let’s consider the use cases once ML at the edge truly can scale: Visualize robotics in a production line that before would be tripped up by a misalignment of products on a shipping conveyor belt that can now optimize and handle any configuration thrown their way. That’s real time inference. Or a consumer packaged goods company wasting supplies and resources on defective products who can now identify the exact point of quality control issues before shipping millions of dollars of product that will not sell. Or a more intelligent medical device that helps identify potential conditions and complements a doctor’s in-the-moment diagnostics ability. In all of these cases there’s no room for cloud latency.

Critical real life ML applications at the edge demand real time inferencing; the latency bar at the edge cannot be met by traditional cloud based approaches. Human beings live at the edge and this is just the beginning for how ML will transform the machines around us.

SiMa now more than ever, is a software company building our own silicon and I couldn’t be more excited about the possibilities Palette Edgematic will afford as our customers, partners and ML community join along for the ride.

Sign up for Palette Edgematic here today!