Edge AI is transforming the way computers interact with the real world. The global Edge Al Software Market is projected to reach USD 4.1 billion by 2028, at a CAGR of 30.5% during the forecast period. Industry expert Gartner predicts 75% of data will be created and processed at the edge by 2025.

Allowing IoT devices to leverage and process sensor data that was previously unutilized due to cost, power or bandwidth limitations, edge AI unlocks the potential for sensors to make smart decisions in real-time without connecting to the Internet. Use cases include but are not limited to, manufacturing optimization, retail personalization, healthcare monitoring, smart city infrastructure and autonomous vehicles.

The challenge of deploying edge AI at scale

The average business cannot easily realize the promise of AI. Deploying edge AI at an enterprise scale requires a high level of expertise and multiple people from different teams -- data scientists, machine learning engineers, hardware engineers -- all working together to test scenarios and environments and to validate that the solution can work on the device in question. As a result, the development cycles for this effort are extremely long, characterized by many iterations and testing to zero in on the right model, using the right hardware, to produce the desired level of performance. Ninety percent of all AI models don’t make it to production, and eighty-five percent of AI projects will deliver erroneous outcomes due to bias in data, algorithms, or the teams responsible for managing them.

Founded in 2018 by Jags Kandasamy and Sek Chai, Latent AI makes it easier, faster, and less complicated for businesses to run AI workloads at the edge. Latent AI’s technology, a suite of software tools designed with AI developer and engineering teams in mind, expedites edge AI development with an all-in-one platform that empowers developers to build secure models ultra-fast, with seamless field or lab updates. The company’s Latent Efficient Inference Platform (LEIP) allows developers to jump-start design with benchmarked configurations, rapidly retrain within a trusted pipeline and deploy secure models that can be monitored in real time. This enables rapid AI application development on lean edge devices at scale, works anywhere, and seamlessly adapts, regardless of framework, OS, architecture, or edge hardware.

Solving AI workflows efficiently at the edge of the network

LEIP allows the compression of neural nets running on any edge AI hardware target in the edge continuum, enabling the Internet of Things. LEIP compresses conventional AI models by 10x without a noticeable accuracy change and allows deployment on inexpensive microcontrollers, DSPs, and other processor cores typically found in edge devices. With many thousands of compelling edge AI applications yet to be developed, Latent AI aims to solve AI workloads efficiently, even at the edge of the networks.

Benefits of using Latent AI’s LEIP include:

• Reduced time to design, train, and deploy models on edge devices.
• Developers and engineers can quickly determine the model size and inference speed for any hardware specification.
• Teams can rapidly prototype and train based on new requirements and train models on less data with the same impact.
• Teams can more easily achieve scalable, repeatable processes across multiple efforts.
• Reducing the need for extensive in-house AI and ML expertise and the need to scale that experience across large development teams.
• Increased performance and maintained accuracy.

Making the leap from defense to private industry

Latent AI found its early traction with one of the most challenging and rigorous of customers -- the US military. Early in 2024, Figure Eight Federal (F8F), Appen’s FOCI-mitigated arm specializing in data enrichment for defense and intelligence applications, announced a strategic partnership with the company. The collaboration is set to transform government sectors by enhancing the entire ML lifecycle, from optimized SME-driven data labelling to delivery at scale of lightweight, ultra-efficient, and adaptive models.

Using LEIP, mission operators have found they can reduce false positives/negatives with AI/ML intelligence, provide feedback in contested environments, increase situational awareness with always-on sensors, upgrade models with new detection signatures and for air, enable fast processing to detect swarm/drones using multimodal sensors, extend flight time and power supply, enable faster detection and response of threats and achieve better target acquisition and recognition.

“The commercial use of edge AI is limited to specific industries such as automotive, manufacturing, and some retail sectors. Edge AI is particularly effective in military applications due to its capability to operate in network-denied environments, and these use cases are currently being refined and scaled for better performance,” explains Latent AI founder Jags Kandasamy. “The trust of the US Department of Defense in our company and our LEIP development platform is a significant validation, paving the way for our growth in the private sector.”

Latent AI has its eye set on the manufacturing sector as a key area for sales focus over the next few years. The smart manufacturing sector is growing 15% year over year. Smart manufacturing operations can leverage edge AI for predictive maintenance, quality control, process optimization, and energy optimization and to enable greater flexibility and agility to adapt lines to respond to shifting consumer demand and changing market conditions.

“Manufacturing is a ripe environment for edge AI and our solution, which reduces latency and development timelines,” says Jags.

He explains that many manufacturers are deploying edge AI today but are doing so within the manufacturing line, often as a hybrid system that is still dependent on processing data in the cloud. This approach comes with latency and cost, which can impact operational efficiency and the bottom line.

To address all use cases from line optimization to quality control and worker safety, manufacturers may find that true-edge AI sensors can be easily added without interruption to power or resource utilization and offer improved monitoring.

“When you process data on the edge, you are far more efficient because you only send what you need back to your main data collection repository for analysis. You aren’t absorbing the cost of a huge pipe to ship data from cameras and sensors to the Cloud. And by processing and using data at the edge, you now have the ability for split-second decision-making - which can make a world of difference in machine performance, line performance, and quality,” says Jags. “And the benefits can cascade across the entire distribution model.”

Jags also points out that even the most advanced smart manufacturer utilizing edge AI must still track and maintain the performance of those models. “That’s not a manufacturer’s core business. It’s a big effort. Is it practical or even possible for a manufacturer to take on that burden?” This is where companies like Latent AI come in with end-to-end MLOps solutions.

Latent AI sells directly to customers and is building out a partner community of integrators to bring its platform to various industry sectors. The company joined Khasm Labs (formerly 5G Open Innovation Labs) Batch 2 cohort in the Fall of 2020. 5G and edge computing are deeply entwined. One enables the other. The speed provided by 5G enables data processing at the edge, and the proliferation of edge computing and use cases justify 5G network expansion.

“Our partnership with Khasm Labs has been instrumental in accelerating Latent AI’s mission to bring the power of edge AI to market. By leveraging their expertise in 5G and edge computing, we've created innovative solutions that address the critical needs of industries undergoing digital transformation,” said Jags.

With its team and platform maturing at pace and the appetite for edge AI rapidly rising, Latent AI is now poised for growth. The company was named a top growth stage AI infrastructure company by Lazard, one of the world’s preeminent financial advisory and asset management firms, and in June of 2024, announced an integration of LEIP with Esri’s ArcGIS software. This integration brings together developer tools with the powerful capabilities of ArcGIS to use AI in drones, sensors, and other edge devices. Latent AI has also joined Esri’s Startup Program, which will help accelerate the development of LEIP to drive geospatial innovation and provide ArcGIS users with real-time, on-device analysis.

“AI will be widely used for both cloud-based applications and edge computing. At Latent AI, we are focused on harnessing the potential of efficient and practical AI for real-world scenarios. We aim to make edge AI accessible to a broader range of devices and empower more developers to integrate edge AI into their applications,” said Jags.