OctaiPipe raises £3.5M in Pre-Series A funding for edge AI platform

Today, OctaiPipe, the end-to-end Edge AI platform for industrial IoT, has announced raising £3 million in pre-Series A funding and a £500,000 grant from Innovate UK. 

Launched in 2022, OctaiPipe provides data scientists and AI engineers working in Critical Infrastructure with a trustworthy end-to-end Federated Learning Operations (FL-Ops) platform. 

The OctaiPipe platform enables users to quickly deploy and automate AI to the Edge and orchestrate and manage distributed machine learning across scalable networks of intelligent IoT devices. 

I spoke to Eric Topham, CEO and co-founder of OctaiPipe, to learn more.

What is federating learning, and why is it so crucial to IoT? 

First, let's do a quick edge computing refresher. 

Edge computing positions intelligence and processing capabilities closer to where the data originates, improving the ability to perform real-time analytics for actionable insights. 

As with scenarios like rugged environments, reducing the amount of data sent to the cloud and between sensors minimises latency and reduces time, energy, and bandwidth expenditures.

Over the last decade, edge computing devices have improved to enable more complex learning on an individual device. But as Topham explained, 

"Systems — particularly automation systems — have a very high sampling frequency. When applied across multiple devices, this results in very high volumes of data. The distributions in that data tend to be heterogeneous. Often, the signals and behaviours you're trying to learn, like anomalies or failures, are rare. 

Therefore, when you try to compute a model on a single edge instance, you perform relatively poorly because you have a low number of relevant events or signals to learn from."

Traditional edge computing moves this data to a central data store. Once the set of model parameters for whatever task you're learning is attained, the data is — usually suboptimally — streamed continuously to the cloud for inference, or you ship the model back down to the edge. 

However, you need to update those models fairly frequently, which means that in the current paradigm, you often ship more data to the cloud, paying for network usage and storage.

Further, the required scale does not exist within a single entity. 

So you end up with use cases where the required scale is limited by barriers to data sharing, such as privacy, IP, and ownership, such as data privacy, data security, and data access rights, especially in the case of mission-critical infrastructure like utilities, defence, and telecommunications. 

Topham explained: 

"You need performance. To get performance, you need scale. Traditional edge computing doesn't give you the scale. The alternative, which is to move data to the cloud, runs into problems of risk and cost at scale. And so you've got a latent value that is just unrealised."

OctaiPipe Federated Learning is a new decentralised approach to training AI models that does not require data exchange between IoT devices and Cloud servers. In Federated Learning, the data on IoT devices is used to train the AI model locally at the Edge, maximising performance and system resilience, increasing data security and radically reducing Cloud data costs.

As Topham shared: 

"The world depends on Critical Infrastructure not to fail but, more than that, to continually improve performance, remain secure and continually become more efficient and sustainable. It's clear that AI has the potential to unlock massive gains in Critical Infrastructure, but only if we can trust that its critical data is secure."

With OctaiPipe, data scientists working in sectors such as Energy, Utilities, Telecoms and Security can, for the first time, use a secure end-to-end platform to design, deploy and manage Federated Learning locally across Edge device networks and at scale. 

Available as a Microsoft Azure, AWS or Private Cloud Platform-as-a-Service (PaaS), the OctaiPipe platform is currently in deployment with over 20 customers and device OEMs.

SuperSeed led the pre-Series A round with Forward Partners, D2, Atlas Venture, Martlet Capital, Gelecek Etki VC and Arm-backed Deeptech Labs are also participating.

Mads Jensen, Managing Partner at SuperSeed, said,

"Critical Infrastructure is a multi-trillion-dollar industry.

Across Energy, Utilities and Telecoms - on-device Federated Learning has the potential to improve performance, reduce failures, enhance security and lead to more efficient, more sustainable services. The OctaiPipe team has already demonstrated significant customer traction, and we are delighted to support them as they scale up to address this important market."

Dr Will Cavendish, Global Digital Services Leader at ARUP, said:

"Water treatment is a complex environment that is expensive for water companies to operate and carries significant regulatory risk, including heavy fines for incorrect treatment.

Federated Learning – including solutions such as OctaPipe's – is an AI technology that can help. It allows continuous learning from multiple and dispersed local data sources, better predicting future challenges. As data from the built environment scales, centralised solutions start to become unmanageable and uneconomic.

So Federated Learning reduces cost and cloud dependence while maintaining model accuracy, security and privacy. Federated Learning also improves system resiliency - meaning there is no downtime risk and systems can remain fully operational in case of an outage or failure."

Miles Kirby, CEO of ARM-backed Deeptech Labs, said,

"At Deeptech Labs, we look for founders addressing global challenges with ground-breaking technology. Eric and the OctaiPipe team are world-leading Federated Learning and Edge compute pioneers.

By applying this technology to Critical Infrastructure as an easy-to-use Platform-as-a-Service, OctaiPipe is helping ensure the services and utilities the world relies on can benefit from the latest advances in AI without incurring the costs and risks of running models on the Cloud."

The money will allow OctaiPipe to develop further its proprietary Federated Learning technology and scale the availability of the OctaiPipe platform for Internet of Things (IoT) dependent critical industries, including Energy, Utilities, Telecoms, Manufacturing and connected device OEMs.

In addition to the funding, Octaipipe has also announced the appointment of Arnaud Lagarde as Chief Revenue Officer. Mr. Lagarde will lead Octaipipe's commercial development. Before joining Octaipipe, Mr Lagarde was the Vice President of Sales at Humanising Autonomy, where he led the global sales efforts and go-to-market initiatives across Automotive, Autonomous Vehicles and smart city solution providers.

Lead image: OctaiPipe. Photo: uncredited.