SiMa.ai is a machine learning (ML) company delivering the industry’s first software-centric purpose-built machine learning software-on-chip (MLSoC) platform for low-powered devices. With push-button performance, the company enables effortless ML deployment and scaling at the embedded edge by allowing customers to address any computer vision problem while achieving 10x better performance at the lowest power. This technology has the potential to transform the embedded edge market by addressing the limitations of legacy technology that have been impeding innovation for decades.
Pulse 2.0 interviewed SiMa.ai founder and CEO Krishna Rangasayee to learn more.
Krishna Rangasayee’s Background
Prior to founding SiMa, Rangasayee was the COO at Groq. Rangasayee said:
“I spent 18 years of my career at Xilinx (now acquired by AMD) where I served as SVP and GM of the overall business, and also led the company’s global sales efforts,growing the business to $2.5 billion in revenue at 70% gross margin while creating the foundation for 10+ quarters of sustained sequential growth and market share expansion. Additionally, I have served on boards of public and private companies.”
Formation Of SiMa
How did the idea for SiMa come together? Rangasayee shared:
“Having spent years in this industry, we recognized that AI innovation was not limited to data centers or smartphones – there is a need for efficient and powerful technology for devices in-between. The semiconductor industry had not yet developed the necessary technology to effectively power the Internet of Things (IoT), industrial IoT, security systems, defense applications, automotive devices, and other machines that exist between smartphones and data centers.”
“We aim to enable AI/ML into these devices and industries, enabling a wide range of applications such as smart factories, assisted driving, automated quality inspection, medical device imaging, and more. Our chips optimize hardware and software to increase performance and save energy in these low-powered devices that are prevalent in various industries being utilized at the edge.”
“Customers don’t want to write new applications or learn new coding languages to support machine learning (ML). Therefore, we developed our ML accelerator to be compatible with multiple open-source and existing legacy applications. This approach allows customers to leverage the technology without needing to acquire new expertise or drastically change existing workflows. We focus on enabling innovation and allowing customers to concentrate on their core competencies while leveraging SiMa’s hardware and software fusion for ML models and computer vision.”
Core Products
What are SiMa’s core products and features? Rangasayee explained:
“MLSoC Platform: MLSoC delivers high-performance machine learning for embedded edge applications. Built on 16nm technology, the MLSoC’s processing system consists of computer vision processors for image pre and post-processing, coupled with dedicated machine learning acceleration and high-performance application processors. Surrounding the real-time intelligent video processing are memory interfaces, communication interfaces, and system management, all connected via a network on chip (NoC).”
“Palette Software: Our Palette software addresses ML developers’ steep learning curve by avoiding the arcane practice of embedded programming. Palette software is a unified suite of tools, functioning much like an ML developer’s familiar cloud equivalent environments, with push-button software commands to create, build and deploy on multiple devices. Palette can manage all dependencies and configuration issues within a container environment while securely communicating to edge devices. This approach still enables ML programmers flexibility to create high-performance solutions without resorting to low-level optimization of the embedded code.
MLSoC Boards:
MLSoC Evaluation Board: Standalone MLSoC evaluation board configuration to explore the full capability of our MLSoC bundled with a one year Palette software license.
MLSoC Dual M.2 Eval Development Kit: Deeply embedded MLSoC embedded module configuration to explore the performance of our MLSoC bundled with carrier card and our Palette software.
HHHL Production Board: PCIe board for edge ML server application, limited software included.
Dual M.2 Production Board: PCIe module for deeply embedded, space-constrained applications, limited software included.”
Evolution Of SiMa’s Technology
How has SiMa’s technology evolved since launching? Rangasayee noted:
“Since launching the company, we’ve demonstrated remarkable success in advancing the state of the art in computer vision and machine learning. We began shipping our Machine Learning Software-on-Chip (MLSoC) Platform kits in January of 2023, and since then customer demand has quickly increased. To keep up with this increasing demand, we’ve grown our team internally, are actively hiring, and were ranked #57 on Forbes Lists of America’s Best Startup Employers this year.”
“We’re currently working with more than 50 market-leading companies across manufacturing, retail, automotive, government and aviation to deploy our highly innovative solution. With the formal launch of our partner program this June, we’ve established several key partnerships with leading companies in the edge marketplace.”
Significant Milestones
What have been some of SiMa’s most significant milestones? Rangasayee highlighted:
“We recently announced additional funding from VentureTechAlliance, bringing our total raise to $200 million. At the same time, we announced that our chips, software and silicon have moved into production. This is a significant milestone for us, demonstrating the value of our technology and trust that our clients have for us. We also launched a partner program with leading vendors in the edge marketplace with e-con Systems, Inventec Corporation, LIPS Corporation, and iWave signing on as initial partners.”
“In April 2023, we set a new industry standard in embedded edge power efficiency with our frames/second/watt measurement system, outperforming some competitors and Nvidia in particular. Our performance in MLPerf’s Inference v3.0 Closed Edge ResNet50 Single Stream Benchmark, which details how much energy is consumed per frame of video, measured in millijoules, demonstrates to the industry that they should take us seriously as competition.”
Funding/Revenue
Upon asking Rangasayee about funding and revenue, he commented:
“We’ve raised $200 million to date, including Series B funding led by Fidelity, with participation from Lip-Bu Tan in May 2022 and a B-1 tranche in October 2022. The funding will help us continue to push the boundaries of what is possible with computer vision and machine learning and accelerate the adoption of its innovative MLSoC Platform.”
Differentiation From The Competition
What differentiates SiMa from its competition? Rangasayee affirmed:
“Where our competitors require customers to hand-code and retrofit their machine learning systems to work on their existing chips, SiMa chips and the software that comes with them are designed to work with customer software from the start – this means AI and ML are baked in from the very beginning, making it an easy solution for customers. We are the machine learning company that has purposefully built the hardware and software platform required for any company to deploy AI and computer vision at the edge on a massive scale.”
Future Company Goals
What are some of SiMa’s future company goals? Rangasayee concluded:
“While we’ve proven we can outperform today’s leading chips in power and speed, we’ve only just scratched the surface of the capabilities and power that our chips and software can achieve. Almost every device currently relies on classic computer vision – in the next decade, everything will be run by machine learning. We focus our attention on partnering with different industries, learning their pain points, and innovating where they need it. On top of that, we’re determined to redirect this industry towards more sustainable solutions. With approximately seven percent of energy consumption coming from data centers and projections showing that rising to up to 12%, the need for sustainable ML alternatives is now and other chip companies will follow to develop more efficient chip solutions.”
