Learn how LEIP accelerates your time to deployment Schedule a demo


The Path to Smarter Edge Devices: The Latent AI Founders’ Story


Interviews with Jags Kandasamy, Co-Founder and CEO and Sek Chai, Co-Founder, and CTO at Latent AI


The age of edge computing is here. The rapid developments in digital and mobile technologies have made edge computing increasingly more prevalent and critical to the success of businesses across a wide range of industries. However, the current process of developing AI applications is complex, training-intensive and heavily resource-constrained, making it complicated for building AI edge solutions.


What if you could simplify AI development workflow and have the freedom to target any hardware infrastructure? This is precisely the solution that Latent AI provides. The new SRI International spin-out venture is changing how AI is built for the edge and shifting the way AI models train for smarter, more efficient and adaptive IoT applications. Smarter edge compute enabled with AI helps increase user experience, improve privacy and reduce dependency on the connectivity.

Latent AI’s mission

Jags: At Latent AI, we are providing tools that integrate with your existing workflow and tool flow to help AI engineers deliver neural net models that are optimized and compressed. We enable those models to be executed efficiently on any chosen platform and edge device. The first product developed out of our solutions platform, the Latent AI Efficient Inference Platform,™ is called LEIP Compress.  It’s a quantization optimizer for edge devices, enabling developers working on edge AI projects to optimize within compute, energy, and memory budget without requiring changes to existing AI/ML infrastructure, processors, or development frameworks.

SRI Ventures’ mission

SRI Ventures invests in deep-tech enabled startups and entrepreneurs by contributing capital, IP, and IQ to make the world safer, healthier, and more productive. SRI Ventures invests in broad technology segments like Robotics, AI/ML, Cyber Security, Computer Vision, Voice/NLP, EdTech, and Healthcare.

How the startup partnership began

Just before launching Latent AI, Jags’ previous startup, OtoSense, had just been acquired. With some new ideas already in mind, Jags wasted little time putting the Latent AI concept into motion.

Jags: I am an entrepreneur, and to bring an idea to life, I knew I needed to find a technical counterpart. When I was ready to move forward, I got in touch with an old friend, Raghu Madabushi, who is an investor at SRI International Ventures. SRI is not only a technical powerhouse, but they also empower entrepreneurs like me to turn bold visions into real-world products. When Raghu introduced me to the Entrepreneur in Residence (EIR) Program, I knew we were forming a great, collaborative partnership.

Raghu: At SRI Ventures, we invest not only in big ideas, but also the entrepreneurs behind the concepts. The EIR program is built to empower entrepreneurs like Jags to launch their next venture. Jags and I have known each other for a while, so I was familiar with his experience, drive, and work ethic, and I knew a collaboration between Jags and SRI Ventures would be a good fit. We work with entrepreneurs throughout the entire early-stage startup lifecycle. I was confident that we could be a partner and catalyst in Jags’ journey.

Jags joined the SRI EIR program, and during a brainstorming session about the technical projects happening at SRI that he had his revelation. It was also here that Jags was first made aware of the technology that Sek Chai had developed at SRI. After hearing about Sek and the efficient edge computing system he was developing at SRI, Jags and Raghu hopped on a plane the next day to meet with Sek. Six hours later, Jags had his technical co-founder. 

Executing on the idea with technology

There is always a huge technology risk a startup must take when launching, especially a deep-tech focused startup. Being able to reduce some of this risk at the early stages can be an essential ingredient for success.

Raghu: This is an area where SRI excels. We are known for creating industry-launching technologies, and these proven technology solutions have the potential to be a key differentiator for a startup’s success. Our diverse portfolio of advanced technology IP has already been tested, validated, and is ready to be brought to life, so this gives startups a big leg up when going to market.

The technology is what attracted Jags to SRI and Sek, who at that time was a Technical Director in the Information and Computer Sciences division at SRI, and an established authority on computer vision and embedded systems.

Sek: SRI is known for its cutting-edge technical work and ability to test and iterate. My team and I had already been working on advanced AI algorithms and low-power edge computing for more than 5 years under funding by the Defense Advanced Research Projects Agency (DARPA) and Department of Defense (DoD). Because of this, we had already established the basic scientific principles for the technology and had provable benchmarks and prototypes. We were ready to go and had been merely waiting for the right business case.

That’s where the Sek and Jags put their heads together and started looking at the market needs and what this technology could solve.

Unlocking the complex AI and edge computing market

It’s astounding to note that over the last two years alone, 90 percent of all data in the world has been generated. At our current rate, we’re producing over 2.5 quintillion bytes of data per day. By 2020, it is estimated that 1.7MB of data will be created every second for every person on Earth.

Given this massive global data generation problem, Jags and Sek recognized the limitations of the cloud in terms of privacy, latency, and bandwidth in processing data since they have both worked in the embedded systems and computing space for decades.

Sek: Robust and efficient AI is still a coveted industry with many compelling user benefits. However, training AI models is already a complicated process, and training for the edge is even more difficult with resource constraints. Leveraging computational neuroscience principles, in particular, the neocortex, we built new algorithms to more efficiently train neural models and to achieve a new level of performance for AI systems that can computationally adapt to changing environments.

Sek: From a technology perspective, we are creating a new wave of ultra-efficient edge AI, called Adaptive AI™, that is a fundamental shift in the way AI trains and how current and future computing needs are determined dynamically. Such technology will improve robustness, efficiency, and the agility of future AI systems. We are enabling edge AI solutions with instant access, no matter the connectivity or situation.

Jags and Sek saw a disruptive opportunity to open up edge AI development to a broader community of developers constrained by a broken workflow.

Jags: We were surprised to learn there were no other companies specifically focused on helping developers work more efficiently in transitioning to the edge. Instead, many startups are prescribing their company-specific, one-size-fits-all models to address all application domains, and there are many others, merely putting out academic reports and papers. For Latent AI, we had a working solution, coupled with research backing, to help developers improve their productivity and time-to-market. We realized a huge opportunity and that a broader addressable market can be opened up.

Jags continues, “I think of it this way. If AI is the goldrush, everyone is trying to secure the land and pickaxe themselves into the rush. There needs to be a supplier of those pickaxes to address how technically sharp and optimized those tools are to help you mine more efficiently. Latent AI is formed to ensure the ‘mining’ process is efficient, and we provide a software tool that integrates within existing workflows, i.e., pickaxes, to be executed efficiently on any chosen platform and edge device.”

Silicon Valley attention: The turning point

With the vision in place, the technology identified, and their “This is going to work!” moment, things started to take off. Momentum began to build for Jags and Sek with interest early on from investors. These initial interests gave them the belief to make the company a reality, but the big turning point happened months later at an SRI Ventures nVentions Day, an event bringing together some of the top SRI researchers, select SRI startups and top-tier venture capitalists. SRI hosted 10 of Silicon Valley’s leading venture capitalists for SRI portfolio companies to share their visions and story. Jags presented the Latent AI story, and this presentation resulted in their $3.5M seed round of funding, led by Steve Jurvetson at Future Ventures, along with other industry leaders such as Perot Jain and Gravity Ranch.

The vision of AI in edge computing: Capitalizing on the promise

Like all startups, there were challenges to overcome, and there are surely more to come. Today, Latent AI has launched with its first product and is already working with the top players in high-end consumer electronics, AR/VR, gaming, auto, drones, and robotics. The goal is to target pilot projects in the coming months, identify the best product/market fit, and figure how to scale from there.

Jags: We have just started scratching the surface of what is possible with Latent AI.  We are enabling customers and enterprises to maximize the benefits of AI in verticals and use cases that were unthinkable before. It is an exciting time, and we are looking forward to paving the path for a smarter edge.