Raj, is our Principal and Chief Technology Evangelist in Connected Devices & IoT, bringing in deep rooted expertise across the ecosystem of practices. Raj drives innovative solutions, frameworks, architectural implementations for our enterprise customers and independent software vendors in areas spanning Artificial Intelligence, Embedded Systems, Internet of Things, IIoT…anything to do with chips and intelligence, lead across our devices, intelligence, analytics and mobility practices.
With that said, let me also share some more perspectives. In fact, at least 50% of enterprise applications in production could be IoT-enabled by 2024, tells Gartner.
So that’s a very interesting world that we’re living in. Another study by Garner tells that 80% of these enterprises and emerging technologies would have AI foundation.
Lastly, we’re looking at a stage where by 2025 enterprise organizations would spend over $10 billion dollars on deployment and maintenance of IT and non-IT devices, supporting everything that can be sensored and monitored - now this is a study by IDC.
With that I take this opportunity to welcome Raj to share, what have been his experiences and how he thinks the platform-led approach at Innominds is truly going to 'Power the Digital Next' in the age of sensors and beyond.
Hey, Raj! So, what do you think are the top trends that you see in the Connected Enterprise and the IoT world?
Raj: Thank you! The first big trend we’re seeing is to bring about something called ‘Digital Sixth Sense.’ This is essentially – IoT, sensor fabrics, acoustic video, environmental sensor radar…time, people, process - technology to drive data insights and subsequently Actionable Intelligence. So, the first trend is, how do you bring about Actionable Intelligence? The second trend – Actionable Intelligence, is great when you have every part of your enterprise connected. So, that could be your backend system, your servers, your data centers, your environmental sensors and solutions for carbon footprints. Many companies have announced a ‘Zero Carbon Footprint.’ And that’s what IoT allows them to do. So that’s been the biggest trend of late.
Sai: So, I hear that, it’s about Actionable Intelligence and what you call the ‘Sixth Sense of Digital’, right! So, what do you think in those cases, could Enterprises and ISVs tap-in by leveraging Innominds expertise in terms of its integrated, platform-led play in the devices, apps and analytics context?
Raj: The IoT universe is complex. Interoperability is a challenge. Many different protocols, different standards, lots of radio technologies - all vying for the same space. And this fragmentation leads to confusion within the CIO and CTO community. So, adoption then takes longer. Fragmentation for services and technology companies like us - that’s what we drive on. One of the things we are good at is to dispel the notion that all of these disparate technologies are really complex to implement. We simplify that, we take a consultative approach to walk a CIO or CTO through simple essentials of bringing smaller subsystems with actionable insights.
The entire digital journey of a connected enterprise can get very complex and is difficult to solve in a short span of time. So, if you want to see a measurable progress, we take simple sub systems, simple things like ‘asset health, asset visibility and asset utilization'. These are simple problems that IoT has solved and that’s our approach – primary, creative approach to some very measurable KPIs for the connected network.
Sai: So, what do you think could be the reason, behind enterprises today integrating sensors, data and devices together? Why do you think they’ve been doing that?
Raj: My personal opinion is that, for enterprises it is a distinct competitive disadvantage without using IoT as a technology strategy to improve their performance. Wall Street expects companies to adopt technology, adopt the digital next to bring about revenue growth, to bring about margin growth and of course, profit growth. Tech firms like Amazon, Google, all of them that are tech-enabled - their strategies and platform-led approaches, they’re setting the bar so high, that traditional brick and mortar companies that don’t have those insights into their operations were significantly penalized by the Wall Street and consequently they have to catch up with these technologies to help become really bright.
Sai: So, it’s basically a perform-apparish kind of a scenario that we’re looking at?
Raj: Absolutely, yeah! That’s imperative; it’s not a choice.
Sai: So, Raj in that context, right! Why do you think Telematics is a very interesting and exciting space? Infact, I’ve heard you talk a lot of Telematics use cases, particularly around the Connected Car market, where do you think this whole Telematics play is coming?
Raj: Telematics is one of the reasons that Tesla, which was an outlier has outperformed every major automotive, in terms of revenue growth, units of cars being made and in terms of the market cap that they’ve achieved. They’ve outperformed in number of ways.
The Connected Car, much like a smartphone is now a software.. an API. That notion, that a car is programmable, that you can orchestrate the performance of the car’s health and to provide a very differentiated user experience. Well, Tesla they’ve set the bar and now every automotive OEM are now playing catch up. They’re trying to bring that kind of Connected Car, Telematics portfolios into their mainstream cars, which is why I’m very excited about Telematics’ market space.
Sai: So, essentially can you talk about Innominds capabilities in the Automotive IoT space?
Raj: The Automotive IoT space spans two cars. The first is the Telematics Control Unit (TCU). This is essentially, the ingress and egress into the cloud.
Raj: The TCU orchestrates all of the communication from the automotive OEM into all of the individual ECUs in a car. As you know, modern cars have uploads of 70 to 80 TCUs.
Raj: And so, these TCUs think of them as small sensors, it has a little bit of memory, has a bit of compute and so the TCU has the rule and apolicy engine, which manages when and how an ECU gets updated. And so, building the TCU hardware, building the software stack on TCU and creating cloud instances where automotive OEMs can leverage the hardware.
Raj: Our job is to abstract the hardware, create a true API-based platform that cloud application can then leverage. So, that’s the first thing that we do in Innominds. The second, is, the car’s brain is like a big processor much like the ones that you see in your smart phones. So, you have a general-purpose CPU, you have a DSP, a GPU, you have lot of audio-video codecs and you have connectivity like Wi-Fi in there. So, this big block chip now powers two things- the digital cluster and the instrument cluster. Most modern cars have lot of instrument panels. It is a fully digital instrument panel. And then you have the in-vehicle infotainment system. So, the new strategy is to have one chip, which drives the brains of the car. It has a hypervisor and one of the operating systems is a real-time operating system, which powers the front-facing camera…near-instantaneous response and then the other one is a high-level operating system - Automotive Grade Linux or what we call AGL or Android Auto which essentially is the application rendering interface in the car. So, in both these portfolios, Innominds has the reference design based on the Qualcomm chipsets for the TCU. On the digital cluster and the AVI, we have a portfolio of chipsets. Some are powered by Qualcomm, some by NXP (IMX) power client, and some by the Intel power client
Sai: In that sense, what would be the five areas, where Innominds could help companies leverage the power of its overall Device Engineering, IoT, AI, Analytics capabilities, essentially to 'Power the Digital Next' - what could be those areas that where we can go to market today - like we’re going and we’re selling? What are those five things?
Raj: The first one is the familiarity that we have with System on Chips. They’re complex, they are powering the edge - System software that is precision-engineered to extract every available mix in the hardware. Autonomous driving for instance, is a very compute intensive. Unless you can shift the computational blocks into the DSP, into the GPU and free up the general purpose CPU, then look at the thermal management, heat dissipation power, leaks so these are complex systems issues, that's something we’re very good at.
Raj: And we have been doing tuning, board support packages, system software images for number of silicon families so that’s one thing. And the second is creating a very well-designed hardware abstraction layer. This is so that Application Developers can then use and share in APIs, to build applications. We don’t expect them to know system calls and intricacies of system’s programs. The universe of API/SDK based developers is significantly higher. So, creating that abstraction layer, that’s the next thing we’ll do and the third one is Automotive Performance and Automotive Stability. In terms of software it is significantly higher than Consumer or Industrial IoT appliances. So, the intensity of testing, rigor in writing the test cases in getting the systems do 120% test coverage – that’s another area that has been of tremendous impact and consequently Innominds you know has matured its Cloud Engineering capabilities in automotive system software.
Sai: Essentially, it’s about silicon…it’s about design capabilities…about creating those abstractions and then the entire ecosystem around this whole thing?
Raj: And the next is of course, with all these computes & smarts, the expectation is that you drive Neural Networks and you drive Deep Learning and Machine Learning. So, we’re inferencing at the edge. We have it in many cars, maybe we have 6 to 8 high speed cameras, full HD cameras coming into the car and Edge Detection, Object Detection, Number Plate Detection, Lane Centering - all of these are functional safety and driver safety related issues. So, the algorithms that are developed are trained in the cloud and deployed at the edge. The tool chain that is required to build it - Deep Learning, Frameworks – TensorFlow, Caffe and many other proprietary Deep Learning frameworks…those have to be plugged into the silicon, it has to be plugged into the other hardware pipelines that the chipset has. That’s another critical component delivering…you know better performance. And ADAS, functional safety and autonomous driving that calls for significant computational intensity in the car and that’s another area where heterogenous computing - the ability to leverage GPUs and DSPs will come into picture.
Sai: That’s quite exhaustive, I must say! In that realm, it’s all about at the end of the day, the engineers who are working on that. So, in today’s context, what could be the main skillsets that an engineer needs to have so that he or she would feel more relevant in the entire AI and IoT sphere?
Raj: Broadly I look at the engineering community as something we call a ‘User Space’ and there’s something called the ‘Kernel Space.’ The Kernel Space is all C C++, Embedded Systems programming, the ability to work with registers and GPIOs. That is one skillset that is extremely useful in making these kinds of system. In the User Space, clearly automotive universe is coming together around Android Auto. So, lot of the Java, J2ME skills; because Android operates Java version on machine so, Java programming skills are relevant to bearing the User Space applications.
Raj: Python is another programming language for developing all the Neural Networks, integrating all of your algorithms to do the inferencing at the edge.
Sai: Sure! Well, the most important thing - would you be able to explain and share what were your experiences working with one of the large customers and partners, where you have been driving the value creation from hardware to software stack and the whole impact that it has created in with the infusion of so many devices. In a sense, you could also explain, some of the recent most exciting customer experiences that you’ve been driving at Innominds and some of those strategic partnerships that you think are significantly creating disruption in the market?
Raj: As a Product Engineering services firm, the expectation is that, we have our platforms to onboard our customers and that is by far the expectation that our customer has that from concept to MVP it is a six weeks hike.
Raj: And so, unless there are some assets and platforms that we can bring to the table; it is very difficult to engage these large OEMs and that has been my experience. In each of these practices, whether it is the devices practice, or applications, or analytics practices, we’ve been at the forefront of building platforms. For instance, in the devices’ business unit, we got the Kiteboard.
Raj: This is much like a Raspberry Pi with Arduino. It’s a proprietary, reference building development kit. So, any customers that come in and stage their application, the system software on our dev kit, we quickly turnaround a couple of use cases and show them a measurable progress to meeting an MVP. In the applications space, we have Microservices, we have few clues that aid in the lift and shift strategy and help scale the applications. So, that’s another investment we’ve made in applications. In Analytics, our wonderful tool called iFusion, we’ll get into it and do a little deep dive later but, these are the three assets that we’ve invested in. These are what we use to onboard large customers. And their experience is essentially, they get wowed! Because you have something to show, bring to the table and quickly turnaround to an MVP and maybe to sprint cycles.
Sai: As we’re coming to the end of it, could you share some use cases that you’ve solved for some of the large customers; you don’t have to necessarily tell who they are
Raj: Smart Manufacturing, Asset Tracking, Asset Management, Asset Inventory. So, these are the areas where you need complex gateways. These are cellular on the egress and egress interfaces, and BLE, PIE, Mesh, Wi-Fi 6 which is a Wi-Fi Mesh, these are all radios that essentially allow smart sensors to connect into the gateways. So, that’s one instance where we ‘ve seen good success. That essentially is, how to bring about multi-radios. We have Zigbee, we have Bluetooth, you have cellular to the whole antenna design, diversity, GPS positioning, location systems, indoor GPS. So, all of these are complicated radio engineering that we have been able to solve in the system software, and that’s one successful product launch that we’ve had.
The second is Trucking. Commercial trucks and fleet, where it’s not just positioned there but, it is also engineering and diagnostics of the vehicle. They also want Fleet Management, Route Management and Route Planning. So, there’s whole bunch of algorithm like cloud-based services that can be leveraged out of that TCU…that’s a segment we have seen growing. Another segment that Innominds has done well is Smart Cameras & Surveillance Cameras. These are essentially inline cameras where you can do Contextual Analytics. Aware Platform, Geolocation, Facial Recognition, Object Detection and Emotional Intelligence are couple of product classes that we’ve built good reference designs for and we’ve acquired good customers.
Sai: To end, the last part of the thing – iFusion. I hear that as an analytics platform, AI powered and ML-driven; how you think the iFusion platform plays a role in terms of these large Device Engineering companies and Connected Enterprises, and helps turn some of these insights into reality? Where do you see the platform play of Innominds coming?
Raj: All of these devices produce significant amount of data. Particularly, in a verbose, logging kind of instances, and most of these devices are in that verbose/logging mode for compliance reasons and regulatory reasons. So, sometimes terabytes of data a day is shifted out of these devices and so the ability to ingest large database and having streaming protocols set up to ingest these data rates into a platform is a challenge and without a significant loss of data, having wire-rated ingestion into a data lake, and into a platform…I think that’s important and iFusion solves that – huge ingestion rates of data; it can be unstructured, structured, or in combinations of these data – video, audio, log data - all that can be ingested. So, that’s the first thing. After you ingest the data, you need to manipulate the data. If you are a data scientist your most effective task and tool is to manipulate the data. Then creating a data mart, which is organized in this data in some sort of relational, NoSQL database or in-memory database - these are performance engineering of a data mart. So, iFusion has tools to create the data mart. Once you’ve created the data mart and you organize the data, then it is important to index and create the metadata. So, again tuning to index very quickly in runtime and create all of the indexes and leverage something like elastic search to be able to quickly create a big search string and on and on…that particular data set that you need and retrieve it. So that is again orchestrating performance engineering of metadata in the data mart. Once that’s done you need to be able to leverage data sciences’ Workbench. It is again a set of tools, or a tool chain that allows you to develop a model.
Raj: That allows you to use a core bunch of algorithms to model your data. And that is another very critical function of iFusion. The final one is analytics after you modeled your data
Sai: Post modelling the data…
Raj: Yes, you want to be able to mission it into the iFusion framework, so the combination of the device from the Microservices to iFusion which is the Big Data warehouse and data sciences’ Workbench and that’s what allows us to bring Neural Networks and AI into play…to the edge.
Sai: To the edge…
Sai: Nice, I think it is quite exhaustive, comprehensive sharing I would say - truly a chip to cloud and cognitive story. More importantly, as I look out there are 26.66 Billion IoT devices in 2019 and the research from Statista says that it would surpass 75 billion by 2025. I think some of these capabilities and the insights you shared makes it very very exciting and to truly 'Power the Digital Next', it is inherently imminent to have a comprehensive landscape that can understand Device Engineering, the Analytics around it and the Apps to build with a lot of AI power intelligence to make it a truly Connected Enterprise so that they unlock the value of digital.
Thanks for being with us... for being so insightful today. We’re here on the first edition of this Innocast. Stay tuned, because we’re going to bring more of this.
Not many companies or System Integrators have end-to-end capability of taking a chip, then building a device around it, connecting it to cloud and run analytics on top of it - both Prescriptive and Predictive. At Innominds, we intend to help drive our customers in their Digital Transformation journey. Innominds has strong Product Engineering background in its DNA, which helps accelerate right solutions for its customers. Coupled with our highly integrated Devices, Apps and analytics solution capabilities, we’ve helped global clients in their innovation and business growth.