Let me give you some interesting statistics to understand why it is so important for the enterprises to know - How Autonomous AI could be of help? The leading research from Gartner predicts that “By 2020 AI will become a positive net job motivator, creating 2.3 million jobs and in fact, it’s going to eliminate only 1.8”. You're actually seeing addition of close to a delta of the magnitude of the new jobs going to create are magnifyingly super. AI and ML are actually at the peak of hype in organizations today where they're actually delivering and starting to deliver significant practical benefits to help solve the real problems. More importantly, I&O leaders need to strategically leverage AI as a core accelerant to their digital business initiative. And that's what we strongly believe. More importantly, to substantiate this fact, the AI is going to drive infrastructure decisions and last but not least, I&O leaders need to strategically leverage AI as an acceleration to business, particularly making decisions in place of humans. And that's what we are here today to discuss about, while going deeper into the topic of autonomous AI, I want to set this context by telling the fact that “By 2023, 40 percent of I&O teams will use AI augmented automation in larger enterprises, resulting in super high productivity, with greater agility and stability”.
Last but not the least, “By 2023, computational resources used in Artificial Intelligence will increase by 5x from 2018, making AI as the top category of workloads, driving infrastructure decisions and definitions”. With that, let me not delay anymore. I am welcoming Ravi. Ravi, as I outline, brings in more than 20 years of industry experience. He heads our AI, Data Engineering, Big Data Analytics practices at Innominds. He advises and evangelizes all our AI based thinking solutions and global implementation for customers across the verticals. He has been the core contributor for our iFusion platform at Innominds, which provides self-service analytics to enterprises and startups and its integrated Big Data as a Service platform which is now being re-pivoted to also do AI on the edge.
He's a giant contributor to 3 patent pending algorithms that do pattern matching, feature engineering and pattern search space reduction. Ravi also drives Blockchain initiatives and anything to do with disruptive technologies. He is an Alumni of the prestigious IIT Madras and IIM Calcutta, and he comes from a rare background of investment banking, transcending, number-crunching from his early days of professional career.
Today, I welcome Ravi and have his insights to be shared. Hey Ravi welcome on behalf of Innominds for our audience. It's my pleasure to be hosting you today.
Ravi: Hi Sai. Hello to all the listeners there. It's my pleasure to be here and sharing my thoughts on AI and Autonomous AI and I am looking forward to this conversation.
Sai: Brilliant Ravi, let's get started on this talk show. So, what in your view, is Autonomous AI and your whole vision of how enterprises can be digitally transformed?
Ravi: Good question Sai. Let me answer this in three phases. In the first phase, I think every enterprise realizes that, like you said, the digital frontier is in AI where whether it is the customer experience management, or whether it is operational efficiency, or a new way of handling risk, or even to manage the workforce, etc. they see the power of AI in enabling fact-based decisions, quantitative decisions, etc. However, we see that while the interest towards AI and the investments towards AI has grown by leaps and bounds in the last three to four years, we see that several enterprises still lack the ability to drive large AI programs seamlessly, almost like a driverless fashion where they can articulate the business benefit upfront all the way up to the data that is required, and manage and massage the data to the algorithms orchestrated between the compute environments including the cloud and the edge. So, while the interest is there, while the commitment is there towards AI, we see enterprises having a challenge in actually driving these large programs and getting the value out of that.
Ravi: So that is where the autonomous AI comes in - that it helps enterprises accelerate this journey of AI And does it in an autonomous fashion guiding all the way in every state and ensuring that there is a package consisting of the tooling, the process and the expertise that is required to drive this AI. So that is what we mean by Autonomous AI.
Sai: Okay. So, in that parlance, let me ask you the other fact - why and how you believe data engineering or in fact AI/ML engineering can actually ensure that Autonomous AI becomes a reality for enterprises?
Ravi: Yes. So, if you ask what's the key challenge in the AI you know the expertise is one, the tooling is another, the process and governance is another. But almost everybody asks this question, where is the data for AI and how am I ensuring and governing that this is the right data for AI, you would appreciate that if you know, if you are letting, for example, machines, like the autonomous car or the autonomous vehicle take some of these decisions such as driving decisions, you have to be absolutely sure that you have trained your models, that you are doing the inference on the trained models, on the right data, and that you're managing the data. Right. So, data engineering is such an important step, an important piece to AI.
As the old adage goes, you know, if you train on the wrong data, you will obviously get wrong results.
Ravi: So, data engineering is important. And the other aspect of that is that data engineering is what consumes the maximum time in an AI lifecycle that people spend enormous time in stitching together the data and massaging the data, cleansing the data, transforming, aggregating. And what do you mean by autonomous AI? Then all of these steps can take care of themselves in a self-service fashion quite intelligently, stitching together all those workflows and stages so that the data is right and ready for the AI.
Sai: Ravi, so far it has been absolutely insightful. In the meanwhile, more importantly the question that grappling in most of the companies minds would be how do you actually measure the success of AI initiatives, assuming one has already embarked on the journey, what has been your experience there?
Ravi: Yes. So, the good news there is that AI can be directly attributed to a top-line increase or a bottom-line improvement, or to a new way of de-risking the business. So, we have seen a numerous initiative towards top-line increase. Anything from a customer experience management point of view of getting to reach out to new customers, or expanding to new geographies, etc. all powered by an AI. We see all of these directly being linked to a top-line growth.
We also see several operational AI use cases where say the mission is assisting the humans or the mission in some cases is looking at some of these operational decisions and handling it by itself and driving efficiency. An example of that is - one of our recent customers where we have the drones taking images off a building from multiple views and an Orthomosaic image is created of the facade of the building and an AI algorithm goes into each of those images and tries to see any cracks or any serious damage to the building.
So, this is an inspection that was being conducted visually by a human who was going to each building and doing this. This is now fairly automated all through the use of AI and avoids the travel and makes this digital or virtual. So, we have seen operational efficiency use cases and we also see a third important use case of AI towards risk management. So, whether it is fraud, external fraud or internal fraud or complying to a standard, complying to a regulation such as say PCI or HIPAA, etc. or in the case of the automotive AIS 140 in some geographies like India. We see that AI can scour through millions of data points looking for anomalies, looking for some deep patterns in fraud, looking for deep patterns and anomalies, and then flagging them off preventively so that brands and entities can prevent any serious breach or outbreak of any of these scenarios. That's the third important use case for AI.
Sai: Awesome. So, as I heard stories around precision farming, you know, agriculture, construction, fraud prevention and anomaly detection, that itself shows us the panoramic impact. You know, successfully AI implementations have brought in and all, so I’m able to understand the fact that these are able to be measured so that enterprises actually unlock the value of data.
Now coming to Innominds, I know that you Lead Big Data Engineering and AI practice and you have been doing that for quite a significant amount of time. Tell us in brief, what are those modern engineering capabilities that Innominds today brings to the table in terms of enabling Autonomous AI becoming a reality for global companies.
Ravi: Sure. We've seen this in three ways. We've seen that one of the key things for AI or driving AI is the data and the data management. So, we setup a process and a tooling, to handle the data, whether it is image data, text to voice, social media or it could be videos shot from phones or from any other cameras etc. We have the pipeline where we have the ability to orchestrate the full pipeline of data in terms of pre-processing the error and signal processing, the noise and noise reduction, particularly with respect to grainy data like images or audio or some of these unstructured data scenarios. So, we have the entire data pipeline with some auto ETL capability so that you are able to figure out as to what data treatment should be given to a given source and so on.
So that this can be fed into an AI. We have got significant capabilities in terms of engineering these pipelines. And we also stood up an integrated Big Data platform named iFusion to achieve this. So that's one piece. The second piece is in the algorithmic AI. We recognized that the research is fast and emerging so as we speak there is always a new kid on the block. You know, this is breakthrough science in terms of say translation you know, language translation or in terms of NLP, conversational capability, etc.
We have the framework here for continuously researching on some of these algorithmic AI, working on them and making them scalable and making them enterprise ready. On the algorithmic AI too we created both the expertise as well as the tooling to ensure that we are able to support not just current research and what is available today in terms of the algorithms, but also the future. For that, we adopted an open policy tool that we will be able to interpret in multi environment, multi ecosystem models. Some of the frameworks are ONNX, openvino etc. So, these are all frameworks to transform the learning from one environment to another. Using that, we ensure that the algorithm AI is ready, not for today but also for the future.
Ravi: And the last piece is the AI in the decision making - the decision making framework where you are able to look at risk reward, pay off and put in a feedback loop so that your past predictions can be analyzed. And based on that, the need for training, the need for retraining, the need for getting more data and more variables into the models, etc., can all be auto learned so that the AI then becomes a closed loop system with its right feedback loops and learning all the time.
Sai: Okay, so from the summary I understand the platform led approach at Innominds where the iFusion is the centerpiece of this whole thing. And you have been able to apply algorithmic AI, you have been able to really go and accelerate some of these large scale AI implementations - is that a fair summary of what I heard from you?
Ravi: It is true. If you don't adopt a platform approach, I think it is easy for enterprises to just get lost in arranging these things and ensuring that the data coming in the algorithms is scalable etc. So. Rather than solving the business problem enterprises can easily get lost in just dealing with the data, the size, the volume of the data and grappling with all the technology.
Sai: Awesome. So I understand as part of the iFusion's road map, you're also talking about an iFusion edge. Do you think Autonomous AI can also have its impact while it comes to enterprise businesses typically getting benefited from edge analytics. What's your overall assessment of where the whole edge computing is going and how do you see an iFusion edge helping enterprises leverage its capabilities? And where do you see Data Engineering, solving those problems? Some thoughts around that!
Ravi: Sure. Computing in general has gone through these cycles from disparate decentralized systems to centralized systems, the client server architectures and then the SaaS and now the PaaS and the full Cloud service offerings, etc. But now we see that the reality of the future is that the compute will be distributed, the compute will be on the edge, the compute will be in the middle of the edge and the cloud, the core cloud itself. We see that in many industries, in many use cases, in many business scenarios. It is important that you have the ability to orchestrate the compute in any of these three layers. For many reasons - one, the reason that there are many use cases where the decision through the AI has to have zero latency and cannot afford to be reliant on the cloud for the round trip as well as for the connectivity. You see that scenario in many scenarios. We see piracy dominating, the confidentiality dominating that each individual, each entity wants to govern the data at their end rather than sending it to the cloud. So, for many of these reasons, we see the edge coming into the picture. We anticipate that at least one third of the use cases will leverage the edge processing capability to be able to manage data at the edge, to do predictions at the edge, to learn from the edge and orchestrate the entire thing at the edge.
Sai: Sure, so I understand that and in my brief understanding of this technically sophisticated subject for the benefit of our audience, what we are speaking about is that for autonomous AI to become a reality, you need to understand the sweepstakes that it can do in terms of its strategic intent and then probably measuring the impact. And then from the use cases as far reaching as from precision and smart agriculture to financial fraud prevention, helping companies to predict in decision making. And more importantly, while the computing becomes reality on the edge, you see that there is a significant touch point, so is that the overall approach that Innominds today is taking to the market.
Ravi: So Innominds just to confirm, is a chip to cloud and cognition company. You know, the North Star for Innominds is to bring the cognition at various layers. So, it wants to bring the cognition to the chip, to the device, to the gateway, and at the cloud. We want to create a smart world that smart devices could run smart factory, smart enterprises. You could do transportation in a smart way. So, you want to be in that in a world of driving.
Ravi: And that's where we see that the intersection of the cloud and the edge and the cognition powering both is a significant interest and opportunity for.
Sai: Right. That's what we are calling ‘Powering the Digital Next’ initiatives of global companies. So, before we end the show, a couple of questions.
So, what do you think are the top five skill sets. Because you have a bunch of data scientists, algorithmic practitioners, advanced data science programmers, under your team. I know that you also have data architects. So, can you talk about that from a skill set standpoint why and how this is exciting and what are the challenges you see in terms of skilling and ensuring that the teams are abreast with the latest stuff that is happening? That's very important for our viewers and listeners.
Ravi: Absolutely. Biggest skill set that would be required for delivering this is to believe that decisions have to be made through data and not just through intuition and not just through cumulative experience that all of us have collected over our lifetime. So, this is an important change in the mindset. We've seen this disrupting and upending market after market. The digital marketing today is entirely different from conventional marketing because data powers the digital marketing, how much you invest, where in what channels, in what ads, etc. precisely to which RDBMS is all now governed by data. So similarly, we believe that the database decision making and similarly insurance is other hall of underwriting today is not through the classic underwriting risk and lost probabilities that they have collected over time. But due to use of AI, they're rewriting those lost probabilities and trying to see as to how they should now underwrite. So, we can see the biggest skill that would therefore be required is that you would have to back your decisions through data.
Ravi: And that is the number one skill in my opinion.
Sai: I just want to add to that question in terms of an engineer's skill set. What Tech Stack, what sort of tools, what sort of programming languages, should someone be equipped with?
Ravi: Fortunately, the world of AI has multiple paths and multiple ways to be doing your job. So, there isn't any monopoly of one tool or one stack. But however, having said that what we believe is that a data scientist should have three core skills from a technology point of view and one of which is the ability to manipulate data and for that we believe that SQL - Structure Query Language, which has been there for ages you know for 20, 30 years is such a core skill that without which, they cannot manipulate data.
Secondly, the data scientist should have one programming language. It so happens that Python probably is emerging as the most popular programming language for AI with a lot of libraries and abstractions, etc. And API’s available in Python, but it need not be mandatory like Python. You could as well do it with Java, Scala or maybe more from a research point of view and also C, C++.
And the third technology stack that people should be aware to do really AIX scale is this distributed processing and big data related, whether it is through micro services and the ability to spawn multiple instances of the compute and the processing to be able to scale AI or whether it is say the Spark ecosystem where you are doing in memory functions in a distributed way. So, we believe that any knowledge over distributed processing frameworks will enhance the engineer’s ability to do AIX scale and not just contain it to some research scenarios.
Sai: Got it. Now, last question, though, in terms of the giant go to market that you are foreseeing. I know that there are advanced pilot initiatives you are driving with global large companies, predominantly If I classify. Could you tell us? How, iFusion edge can do Streaming analytics integration or probably a iFusion edge, in the area of AI Ops and Data Ops where do you see in some of these use cases.
Ravi: Yes. So, we see that a lot of the IoT world comprising these devices that are all connected in a way they will have to be managed, maintained and made as a service and being agile, etc. And AI can drive this. And that's exactly what we mean by an AI Ops to be able to drive operations proactively through AI and make high availability and as a service real.
Ravi: So, we see that as a predominant use case in telecom, healthcare, transportation. You know, every telematics device that is inside a truck is telling everything about the truck, including where the truck is, what the engine condition is, how the driver is, a sort of say driving the truck and so on. AI Ops becomes then a reality for the fleet operator to be able to see the truck, see its condition and do things like dynamic routing and write and change road paths. You know options are optimized for some aspect of the fleet management and so on.
So, we believe that these are important and for that we have partnered with some of the leading technology companies that will bring different aspects of AI Ops to us. For example, one of our partners or a customer has a platform for service management, so their core solution has the ability to track assets and track enterprise assets particularly and manage the services around the assets. So, we partner with such a player to ensure that we can bring AI into service management and convert that as to do an AI Ops.
Ravi: Similarly, another partner of ours, who is big time into enterprise data analytics and enterprise security, we're enabling the IoT side for them. You know to be able to collect the logs and other information from IoT devices, apply our AI and see that this physical world of IoT devices which are not so much within the proxy or within the firewall of the enterprises are equally protected as something that is within the proxy server and within the firewall.
Sai: Awesome. Ravi, it has been a really insightful session. I must say with so many deeper perspectives that you're sharing, now the world is obviously in the brink of the most dangerous pandemic that we've all experienced. So, in your view is there any singular aspect of implementation, that AI could be looked at probably as a sweeping statement that this is something the implication of AI that, you know, the post pandemic era could also witness immediately, are there any benefits you can talk about?
Ravi: Absolutely. Here is my one liner on that. The vaccine that is going to be developed for the COVID19 will come out of an AI algorithm that tries the genome sequence of the virus, as we know, against all known drug combinations and all known antigen antibody combinations. Right.
Ravi: And tells us as to what will work against this. So while that is the big ticket AI is helping us to deal with this crisis. For example, much of much of the scenarios of shelter at home and quarantines etc., have all come out of prediction models that will tell us as to how the spread of virus is going and how we can actually slowdown that spread and protect ourselves.
So, we see AI being a big player in this in the COVID 19 scenario and helping us stay safer and helping us prepare and rebound stronger through this crisis.
Sai: Thank you Ravi. You stay safe and I'm sure every one of us. Let's maintain the social distancing and ensure that we follow whatever it takes to defeat this pandemic. But once, I'm sure, we bring back to normalcy artificial intelligence in its current and future form, I'm very optimistic that it could enable a truly autonomous enterprise and we will look forward to having continuous insights from you in the sessions that we continue to bring. At Innominds, our endeavor is to ‘Power the Digital Next’ companies through a smart integrated approach of our capabilities around everything - Software Product Engineering, Device and Embedded Engineering and more importantly, AI, Analytics and Data Engineering.
Thank you for being today on the show. Innocast will continue to bring significant thought leadership insights that I'm sure would find more useful anecdotes for all of you to take back. Thank you, Ravi
Ravi: Thank you all. Thank you all. Stay safe and stay healthy.