AITech Interview with Kenta Yasukawa, Co- Founder and Chief Technology Officer at Soracom

Discover the technological advancements shaping the IoT industry in an interview with Kenta Yasukawa, CTO at Soracom. Get unique perspectives from a seasoned industry leader.

Kenta, it would be greatly appreciated if you could provide us with insights into your professional trajectory and background that culminated in your position as the Co-Founder and Chief Technology Officer at Soracom.

In 2010, I was a researcher at Ericsson working on Connected Home, Connected Car and similar early-days IoT projects. I was drawing diagrams with fluffy clouds in the middle saying all the intelligent decisions would be made and things would get smart and collaborate with each other once they are connected to this new cloud. But back then the only available technologies for that purpose were rule-based engines, inference based on ontology, which did not have enough potential to be the intelligence to achieve a true Internet of Things vision.  

I felt there was potential in cloud technologies, but I didn’t fully know yet what the cloud could offer. So, I joined AWS as a solutions architect to find out. I worked with various customers to architect systems in the cloud and apply AWS best practices. That made me think, that by applying cloud technologies and best practices, any system can be made more reliable, scalable and available. It should be possible to build telecom infrastructure on top of cloud and it should enable a highly scalable connectivity platform.

I shared the idea with Ken Tamagawa, my CEO and Cofounder. He believed in the idea and we started to seek a way to execute and along the way met Dan Funato, my COO and Cofounder. We founded Soracom and I led the reinvention of telecom infrastructure on top of AWS cloud.

Leveraging the cloud-native telecom infrastructure, we have started a smart connectivity platform that can offload customers’ undifferentiated common heavy liftings in their IoT journeys and accelerate their time to market so we can achieve a truly connected world together.

The recent announcement  introduces Soracom’s new services that leverage Generative AI (GenAI) for IoT connectivity. Could you explain how Generative AI fits into the IoT ecosystem and what advantages it brings to IoT deployments?

GenAI has tremendous potential in IoT deployments. Besides adding natural language interfaces to IoT applications, GenAI applications using a Large Language Model (LLM) in particular has potential to be used for data analytics and decision making.

For example, we have tested ChatGPT to analyze time series data received from IoT sensors and trackers and confirmed it can provide insight about data as if you have a data scientist dedicated to you. By providing data and asking questions such as “What does this data mean?” and “What trend or outliers do you see in the data?”, an AI can answer in a natural language that you speak. We realized the potential and integrated GenAI to our time series data storage service, Soracom Harvest. The feature is called Soracom Harvest Intelligence and available to anyone as a public beta. An AI based data analytics is just one-click away

As in the example, GenAI can be a glue between people and data, and help them understand data. This can help people look deeper into a particular time period, detect an event and take action. If it has to be done by humans, it’d be cost prohibitive and not be scalable, but with GenAI, things can be automated and scalable.

Soracom Relay allows customers to use existing RTSP/RTP-compatible cameras for audio and video data transmission. Can you discuss how this feature enhances IoT deployments, especially in terms of computer vision and video analytics?

There are many cameras available today that claim to be “connected to the cloud,” but most of them are tightly integrated into their vendors’ vertical cloud applications and require wholesale replacement of existing hardware to take advantage of these capabilities.

RTSP is a standard protocol already widely used in various IP camera products, making them easy to integrate within a tech stack but typically only in single building on-site CCTV deployments. The ability to securely connect RTSP cameras securely to the cloud opens the door to advanced monitoring and analysis capabilities without needing to change your entire set of cameras to shift to cloud-based video processing.

Soracom Relay enables a complete new set of potential opportunities to create value from existing view/record/replay cameras over to new architectures that connect those same video streams to cloud-based processing. Single-site installations typically use disk-based recorders that implement RTSP/RTP connections so that video streams can be viewed locally, sometimes with additional proprietary cloud features for simple remote-view functionality.

With Relay connecting cameras to Amazon Kinesis Video Streams, we encapsulate the overhead of implementing the RSTP/RTP protocol into the camera’s Soracom connection and let the customer shift to an AWS cloud compute architecture to process and create valuable business-centric insights.

Security is a significant concern when dealing with data transmission, especially in IoT. How does Soracom ensure the security of audio and video data transmitted through Soracom Relay?

The security and privacy of customer data is the highest priority in the Soracom platform. We have multiple layers of security implemented in our platform and also offer services for our customers to build a secure infrastructure that supports the needs of  their particular IoT fleet.

Soracom Relay is an ideal match for the secure architecture at the heart of the Soracom IoT Platform. When we saw this use case we immediately knew that we could mitigate risks and concerns associated with RTSP/RTP connected cameras while opening up new revenue possibilities for customers  by linking cameras that traditionally only have a LAN connection directly to AWS’s video streaming services.

RTP video streams are not encrypted and can be dismissed as a source of IoT data despite the very large numbers of devices deployed in LAN environments. Similarly, the RTSP servers these cameras use have often been implemented with poor administration account credentials.

When IoT devices use Soracom connectivity they benefit from a fully encrypted link for all traffic between the devices and our connectivity platform. In the case of RTSP/RTP cameras, that means that the account login process is completely locked down to the Soracom account and moved out of reach from bad actors. The valuable video streams become tamper free, ensuring both that the stream is trustable and at the same time unavailable for others to access.

The integration of RTSP/RTP and AWS KVS provides a trusted source of data directly to the AWS cloud compute environment without the overhead or risk of trying to develop secure solutions to resolve the camera’s account or protocol shortcomings.

Soracom Query enables customers to mine IoT device data using SQL queries. How does this capability eliminate the need for users to set up their own servers or storage for data analysis?

Soracom Query is our response to the evolving needs of IoT data analysis and data pipeline – comprehensive, yet remarkably streamlined.

Understanding the dynamics of IoT device data analysis, we recognized the gaps our customers were facing. Traditional visualization tools offer a glimpse into the data, but leave deeper insights unexplored. The increasing volume and complexity of IoT data underscored the need for robust cloud capabilities. Moreover, data synchronization with warehouses was an intricate task, requiring niche expertise.

Our existing solution, called Soracom Harvest, already offers basic serverless data storage and visualization, but with limited capability for more complex and voluminous data sets. And though we provided an avenue to set up data warehouses, this demanded significant technical know-how from our customers.

Enter Soracom Query. This solution blends the technical capability of Soracom’s cloud-native IoT connectivity platform with user-centered design. Soracom Query provides a fully managed data warehouse backed by Amazon Redshift. This lets us ensure that large data sets can be stored efficiently, while also offering customers the liberty to choose their preferred BI tools, including Tableau, PowerBI, and Quicksight. The beauty of Soracom Query is its simplicity. Customers can dive deep into their massive IoT data without the concerns of managing, maintaining, or understanding the intricacies of a data warehouse.

The managed data warehouse capability of Soracom Query is mentioned. How does this capability handle large IoT datasets and supports analytical queries?

So far we have provided Soracom Harvest for storing time-series data and            Soracom Lagoon to create a visual dashboard for data stored in Soracom Harvest. This combination of services lets customers set up a realtime dashboard and alert functions. However, when it comes to running an analytical query e.g. for the entire history of stored data collected from the entire fleet, the existing services we have offered were not designed for that.

Our customers could have set up their own data warehouse to overcome the challenge and integrate using our protocol adaptation or cloud function services. However, it’s still a heavy lift on the customer side because synchronizing data to a data warehouse requires knowledge and expertise by itself. We recognized that capturing and organizing the data from each device is only the first step in creating data-sets that align with the requirements of existing business systems.

Therefore, we came up with Soracom Query, a data warehouse integrated with the Soracom platform and managed by Soracom, leveraging the benefits of Redshift’s fully distributed columnar-based data layout. Even if customers store PBs of data, we can store them in a storage that applies different compression algorithms to columns of data to reduce storage space and increase data loading efficiency, and they can run a complicated analytical query by using SQL. They can also use a BI tool of their choice to run those queries and analyze data. Since data synchronization and maintenance are fully managed by Soracom, customers can focus on analyzing their data instead of undifferentiated heavy lifting in IoT data analytics.

Soracom Harvest Data Intelligence focuses on enhancing data storage and visualization by applying Generative AI for analysis. How does this service assist users in identifying trends, patterns, outliers, and abnormalities in time series data?

We created Soracom Harvest Data to manage time-series data storage so that our customers can collect data from their IoT devices and run time-series queries to detect events, trends, anomalies, patterns etc. A device using Soracom connectivity can send data without having to use an SDK or credentials to authenticate because the service endpoint authenticates the device by using its connectivity service ID (e.g. ICCID/IMSI in case of cellular) and stores data payload with the ID and timestamp. The simplicity and ease of use have made many customers choose the service as the first entry point for collecting and analyzing data.

Soracom Harvest Data Intelligence builds on Soracom Harvest Data to let our customers bring generative AI functionality to their IoT data, build valuable insights close to the data-source, and empower their systems to work with insights instead of transferring data for analysis further along their architectural chain.

This effectively brings a GPT AI-based data scientist just one click away to every customer. All the user needs to do is to click the “Ask AI” button on the Soracom Harvest Data user console to receive analytical results in their localized natural language. To use the service, the user does not need to think about questions or prompt engineering. We’re providing a set of prompts that our customers can use to get started on this process.

We’ll be keeping the dialogue open to learn how our starter set fits the needs of our customers, how we can tweak the prompts, and how we can grow the set of actions that customers can take. Soracom Harvest Data Intelligence is an extension to our Soracom Harvest Data service that customers can already use to capture the sensor data from their devices, when our customers use the Harvest Data Intelligence prompts they can see the text output of running prompts directly on their Harvest Data while it’s still within the Soracom platform.

Are there plans to expand the capabilities of these services in the future? Can you share any insights into how Soracom envisions the evolution of Generative AI in IoT connectivity?

Another possibility we would like to mention is to apply GenAI to find insights from a large, complicated dataset that involves data from different domains or verticals. Today, most IoT systems are isolated and siloed, and data collected for a system/organization stays in the system/organization. It is a right thing to do for security and private reasons and we, as a platform service provider, encourage that and offer features to achieve such isolated environments. However, when the concept of IoT was born, what people envisioned was a world where everything and everyone was connected and interacted with one another to make the world a better place. Compared to that, what we see today is many instances of Intranets of Things rather than true IoT.

When we aim for a true IoT vision, there are many challenges. Let’s say we have a way to share data securely with only designated parties with appropriate filters to protect privacy. Still, what is missing is an enabler that can take data from various domains and verticals, understand the correlations and derive insights. Back in my research days, there was no good answer on how to implement such an enabler. Of course, there is no way to implement a rule-based engine that handles data with always growing dimensions.  Inferences based on ontology did not open a way forward. GenAI or an LLM has the potential to be such an enabler. It does not have to have a training specific to a particular domain to get started. It can take CSV with attribute names to provide insights. OpenAI Code Interpreter can take a large chunk of data and write code to analyze. By leveraging evolving GenAI technologies, there is a potential to connect dots in data from different domains and draw a conclusion from that, which is beyond our imaginations, e.g. farming sensor reporting dry soil can be an alert for a drink company to fill more drinks in vending machines in the area. We believe we are getting closer to a true IoT vision where things from different domains share data and knowledge and actuate things in the field based on it.

To wrap up, is there anything else you’d like to add regarding the intersection of Generative AI and IoT connectivity, or any key takeaways you’d like our readers to remember from this announcement?

As mentioned in the answer to question 2, GenAI can fill a piece that has been long missing for achieving true IoT vision where everything and everyone is connected and interacts with one another to make the world a better place.  It has been our company’s vision since day one. We are excited to see that we are getting closer to achieve the vision by leveraging evolving GenAI technologies.

As a way to accelerate the progress, we established the IoT x GenAI lab with Matsuo Institute, who shares the same vision as Prof. Matsuo’s lab in Tokyo University. The primary mission of this lab was not just about technological exploration. It was about understanding the pulse of customer needs, diving deep into the ocean of IoT, and surfacing with pearls of innovative solutions. They sought to continuously monitor the evolving dance of GenAI and LLM, deciphering their intricate steps, and choreographing them into the grand ballet of IoT services.

But the lab’s aspirations didn’t stop at research. They envisioned a world where businesses, big and small, could harness the combined might of IoT, GenAI, and LLM. They dreamt of prototypes that could revolutionize industries, of technical verifications that could redefine norms, and of supporting user companies in weaving their own tales of innovation.

Kenta Yasukawa

Kenta Yasukawa, Co- Founder and Chief Technology Officer at Soracom

Kenta Yasukawa is CTO and Co-Founder of Soracom, where he has led deployment of the industry’s most advanced cloud-native telecom platform, designed specifically for the needs of connected devices. Before co-founding Soracom, Kenta served as a Solutions Architect with AWS and conducted research for connected homes and cars at Ericsson Research in Tokyo and Stockholm. Kenta holds a PhD in Engineering from the Tokyo Institute of Technology, with additional studies in Computer Science at Columbia University’s Fu Foundation School of Engineering and Applied Science.

Related posts

AITech Interview with Michelle Songy, Chief Executive Officer, Press Hook

AI TechPark

AITech Interview with Aran Khanna, CEO and Co-founder,

AI TechPark

AITech Interview with Niraj Kumar, CTO, and Co-founder at Datametica

AI TechPark