THE RARE ‘HARD-WEAR’ MAN OF INDIA!

Drawing inspiration from the legacy of India’s former President, Dr. APJ Abdul Kalam, Professor Veezhinathan Kamakoti, the Director of IIT Madras, is actively pursuing Dr. Kalam’s mission of contributing to the nation’s progress.  Prof. Kamakoti led the research team that developed India’s first indigenous microprocessor, ‘Shakti.’ He has a clear vision of how he intends to enhance the innovation and technological prowess of the country.

Attempting to define him in a few words would be unjust. Prof. Kamakoti’s charisma is evident in his analytical skills, detailed conversations, rational thinking, and vast knowledge.

In an interview with Niloy Banerjee from THE VOLT POST, Professor Kamakoti, affectionately dubbed the “Hard Wear Man” by his interviewer, discusses the semiconductor industry in India, provides insight into ESG [Environmental, Social and Governance] matters, and shares his thoughts on ‘Chips’ AI & BHARAT GPT in his distinctive and engaging style. What follows is the edited version of the interview from the popular video – the language has been edited for clarity and brevity without altering the context and content. Full Video Interaction is available to watch at #CCTV (COFFEE CLUB WITH THE VOLT), 

Kindly share an anecdote behind developing the first indigenous Indian industrial grade processor named, “Shakti”?

Around 2011, Dr. APJ Abdul Kalam urged the youth and technocrats of our country to take concrete action rather than abstract endeavors for the Nation’s progress. This meant the creation of something worthy and not just telling stories. Dr. Kalam’s words were an inspiration to us and we are all trying to pursue and fulfil his mission. At the time Dr. Kalam said that, I was working in the area of semiconductors and that’s when we decided to create a microprocessor for the country.

In 2011, we envisioned a digital revolution and recognized the need for a family of microprocessors that could power miniature devices as well as large-scale cloud servers.  We needed to start putting in effort in that direction. We started something that we have sustained for over twelve years. We have successfully made three or four chips and have proven their effectiveness.  We have also induced more than 1000 teams across India to use these processors as a part of a ‘SWADESHI MICROPROCESSOR CHALLENGE’ that was hosted by the ‘Ministry of Electronics and Information Technology’, Government of India. Today, more than five startups that originated from IIT Madras are working in the “SHAKTI” initiative.

Can India nurture hardware-focused semiconductor startups alongside its service-based software startup ecosystem, in line with its vision to establish itself as a semiconductor hub?

Absolutely. I’m sure that we are on the right path. I truly believe that it is not a fairytale anymore but a reality. The pioneer of such programmes, the Digital India RISC-V (DIRV) initiative, was launched by the Honourable Minister, Mr. Rajeev Chandrasekhar. Programmes such as DIRV have the potential to foster semiconductor startups focused on hardware development. Presently, numerous startups are addressing fundamental aspects of microprocessor development. In my view, the microprocessor stands as one of the most intricate electronic circuits humans have ever conceived, developed, and manufactured. In that direction, DIRV can attract several startups that work on different aspects of this chip development and create a viable ecosystem across the country. This will help address different issues and provide solutions.

What are the key security challenges and architectural design complexities in democratizing processor-based SoC design?

I must put on record that Linux, which is an open-source model has been customized by many organizations. Undoubtedly, today it is the most secure operating system because many eyes have looked into it. Vulnerabilities have been caught and people have fixed bugs and issues over a period of time. Even in the rare event of a vulnerability, scrutiny by a vast number of individuals ensures that issues are promptly identified and resolved, unlike closed-source systems. Therefore, I firmly believe that the open-source nature, coupled with the ethical scrutiny from academics and industry experts, enhances its promise.

Drawing upon my 24 years of experience in teaching hardware, I advocate for the promotion of open-source platforms for customization and product in development as the optimal solution for security concerns. This sentiment aligns with the ethos of transparency and collaboration inherent in open-source systems.

It’s worth noting that Berkeley initiated efforts in this direction through the RISC V, with IIT Madras being the third academic institution to join after Berkeley and MIT. Despite initial skepticism, particularly during fundraising in 2017, we remained steadfast in our conviction that openness is paramount for security, whether in software or hardware. Linux has provided us with the confidence to make this assertion proudly.

“I am truly very elated to tell you that like India has a National Flag, National Song, and National Anthem, we now have a National Instruction Set Architecture (ISA). This will democratize security. Furthermore, we have integrated numerous secure features into our microprocessor, the Shakti Platform, ensuring robust architectures with demonstrable security.

Our microprocessor like any traditional microprocessor/conventional microprocessor works across four levels: Machine mode, Hypervisor mode, Supervisor mode and User mode, each with its own privileges. There is a lot of protection that one can assign based on these modes. Now, SecurWeave, an IIT Madras incubated company, has developed a separation kernel, a first of its kind, that executes in the hypervisor mode. This kernel effectively isolates the operating system and the software from the hardware.  Security threats such as Root Kits can be significantly and effectively mitigated. These robust security measures are all open-source. The ‘Shakti Platform’ is instrumental in crafting such secure architectures.

SHAKTI, with such features is put in the open web for people to start using it. People can do their own due research in terms of security etc. Many of them have come back and told us about things that needed to be fixed. We fix them and give it back to them. This process of open collaboration and feedback is pivotal in democratizing access to technology. I strongly believe that this transparency, coupled with democratization, is crucial for fostering the development of a secure, compute infrastructure and promoting the same to startups in the country.

Dr. Kamakoti, the next question is about the Bharat GPT initiative, we want to know your key role in it.

Prof. Ganesh Ramakrishnan, from IIT Bombay is leading the Bharat GPT initiative and other aligning and associated institutes are participating in it. We truly believe that ‘GPT is like a Kid’; it learns and responds based on the input it receives. Therefore, the fundamental input provided determines its future responses. This is like nurturing a child to respond in a certain manner as it grows.

Our role in BharatGPT is to work with IIT Bombay under its distinguished leadership. Our objective is to advance the initiative by developing domain-specific Large Language Models (LLMs). Generally, the aim for these models is to function as ‘Jack of all Trades’, capable of answering queries on a wide range of topics. However, achieving this ambition entails increased power, computation, and storage requirements, resulting in a significantly larger compute stack.

As with computer architecture, there exist two primary types of machines: general-purpose machines and domain-specific architectures, spoken about by Profs. Hennessy and Patterson, who are regarded as pioneers in computer architecture. Domain-specific architectures cater to specific tasks without the need for excessive computing resources. Thus, our focus lies on developing such architectures tailored to specific domains, rather than relying solely on general-purpose computing solutions.

For example, I can have something called an extensible instruction set architecture, where I can take some part of the instruction set and make a processor out of it and the tailor-made compiler will only compile those instructions. Similarly, when it comes to Bharat GPT, we need to focus on domain-specific LLMs, such as those tailored for legislation or agriculture. It is important to avoid confusion with other contexts, languages, or jargon, ensuring that the responses are distinctly Indian, addressing Indian queries.

For example, for farmers, the system should be capable of responding in multiple languages. Users should be able to input a question via an app, and receive answers in their local language. The same applies to Indian healthcare, where Bharat GPT’s approach and role will be crucial.

Being one of the collaborators, IIT Madras will be contributing to one or two domains. It is a collective effort, many institutions are working together and I am sure it will be very successful under the able leadership of IIT Bombay Professor, Ganesh Ramakrishnan.

What do you think of the importance of ESG and Public Policy and amalgamating advanced engineering and research thinking mindsets across it?

ESG is becoming extremely important. You talk to any bank today or see their financial report, they’re all talking about ESG, though ESG is still an evolving concept. For every innovation, we need finance and funding and any process becomes difficult without money. Let’s talk about a bank, Let’s say, I’m sitting in a bank, somebody is coming up and asking for a loan to set up a ‘Metal Company.’ Now imagine the amount of carbon they would emit in manufacturing that particular product. Certain countries may refuse to purchase their products, if they exceed certain regulatory limits, emphasizing that it’s not merely the end product at stake.

Take another example of a “Computer Mouse” in our hand. This mouse has some plastic, then the internal part has a few electronics, levers and there are even screws. These parts are all made by some factory and if the factory does not meet the “Emission Norms,” then the entire mouse would be rejected. This shows how interconnected things are and why it is crucial for countries to meet their commitments to Sustainable Development Goals, which the UN emphasizes and is highlighted in G20 summits.

So now as a bank, when I’m trying to even give a loan, I have to start understanding the ESG from the grassroot level. I cannot suddenly come and superimpose ESG on all the assets that I have created and all the credits that have been availed from the bank. I must start at the beginning. These are very interesting concepts and public policymakers, essentially regulators like RBI, need to further engage in ESG. Then, this whole system must amalgamate. If funding is granted without considering ESG, at some point, the NPA will rise. There should be a public policy, and we need to amalgamate it with some of the understanding that we have of this whole concept.

If we are funding a factory, we have to ensure that it has the engineering capability and that over some time it will meet certain ESG norms. E, S and G are interconnected, so we need to have a strong policy-making system for environmental policies. We would also need a firm governance structure to quickly adapt to the ever-changing environmental changes and consequences. What we perceive as effective environmental policies today might not suffice in the future. Are we equipped to promptly adjust and amend these policies?

So for this entire evaluation of ESG, there needs to be a lot of engineering interventions. For example, today we were evaluating a project on using sewage water in a boiler to produce steam. This steam would then facilitate the process of paper crushing to create high-quality recycled paper.

Recycling paper requires steam, which in turn must be generated from a boiler that heats recycled sewage water. Municipal waste is used to generate heat. Imagine! We take a paper that is dumped and recycle it to create good-quality paper using Municipal waste and sewage water!!!  Now, when I mention ESG in this context, someone seeks funding for this innovation, and it intertwines with engineering concepts. For example, considerations arise about how to recycle efficiently, what type of membrane to use for recyclability, and how to effectively segregate municipal waste. We must also ensure that any boiler employed meets pollution control norms.

Furthermore, how do we execute these plans? If the pollution control requirements change over time, do we have the engineering capability to adapt accordingly? Will these changes make the business sustainable? A lot of questions arise, and answering them requires a critical blend of public policy and engineering.

Moving on to my next point, India’s vibrant dynamic nature poses a challenge in nurturing talent pools aligned with the country’s vision for the future, particularly in research implementation and bringing ideas into final products.

Let me illustrate with an interesting project we are undertaking with the Tamil Nadu Government. When it comes to semiconductors, prioritizing human capacity building is paramount. Investing significantly in developing human skills is our true wealth. My recommendation to all funding agencies is that 80% of project resources should focus on human capacity building, with the remaining 20% allocated to capital. This holds true even for semiconductor projects.

How many people are interested in hardware?

When I joined the Computer Science Department of IIT Madras on July 1 2001, I was the only person working in VLSI hardware, which became very interesting for many people. I still remember more than 70% of B Tech students wanted to do a project with me. Right then came my friend Ravindran who brought in AI. Then, the attraction turned towards AI.

Now I find myself telling people that I’m doing AI hardware, which has turned heads and prompted many to explore the integration of AI into hardware.  Today people are doing AI/ML related jobs because they seem to be very stable and pay handsomely. But, we need hardware to execute these algorithms. All these systems, all these algorithms, and software work on hardware. Now, who will build hardware? This is the question, a very fundamental question.  So, we need people, who are extremely interested in electronics, we need to develop that interest so that they may engage more actively.

Coming back to our work with the Tamil Nadu Government’s School Education Department, we have a very vibrant school Education department in Tamil Nadu.  At IIT Madras, we have developed an in-house electronic kit, which we provide to students from the ninth, tenth, eleventh, and twelfth grades to conduct 25 experiments in electronics. By the time they reach twelfth grade, they will have completed 100 experiments. We anticipate around 100 students participating in each class, totaling 400 students per school. With 252 schools participating, approximately 100,000 children will benefit from this initiative. Assuming 25,000 children complete 100 experiments each year, they will hopefully, develop a keen interest in electronics through hands-on experience.

The experiments are designed to align with their BE Electronics and Communication syllabus for the first three years. Through these experiments, they will grasp basic concepts in electric circuits and related topics, preparing them for potential engineering pursuits in electronics and communication. Assuming that only 20% of the 25,000 students are interested, we will still have 5000 government school children proficient and interested in electronics. Having completed these 100 experiments, they would have gained a solid introduction to electronics. This approach is crucial in advancing innovation—we must ignite their interest, motivate them, and provide opportunities for growth.

It is also very important to help interested students explore further. Keeping this in mind, we have started a BS in Electronic Systems – it is an online course. We also have a BS programme in Data Science where we have 25,000 students currently enrolled.

In BS Electronic Systems course, we have 1,000 students now. These are four-year courses, and the students receive the degree from IIT-MADRAS and they become alumni of IIT Madras.

Another important aspect is encouraging people to patent their ideas. Today, the Government has facilitated so many patents; when I first started my career, the word ‘PATENT’ was completely unheard off.

Last year alone I was granted four patents. Nowadays, the patent disposal process takes 6 to 18 months. AI-based tools help navigate multiple patent databases worldwide, providing insights into patentability and market potential.

There’s a Chandigarh-based company, XLPAT, which has done exemplary work on this and we are using some of these patent search systems. We filed 366 patents during the financial year of 2023-24. We want to push people to own their ideas and then these patents must translate into  products. We at IIT Madras are very clear about this, we have an Entrepreneurship Cell, we have a Centre for Innovation, and multiple incubators. The B.Tech and M.Tech Projects can be explored for patentability.

Now let’s say someone at IIT Madras has an idea, they can come to the ‘Center for Innovation’, which helps them translate this idea into a design. Then they go to ‘NIRMAAN’ which is a pre incubator. They help convert this design into a prototype. We have incubators in IIT MADRAS, we have an incubator for ‘Cyber-Physical Systems’ IIT Madras Pravartak Foundation, health technology innovation, rural technology based incubator, and a bio incubator. A sustainability incubator will also start functioning soon as will another incubator dedicated to financial technology.

These incubators will basically convert that prototype into a startup.

The Gopala Krishnan Deshpande Centre, established by two distinguished alumni, mentors and supports startups, with the aspiration of transforming them into unicorns. As the top-ranked institution in the NIRF for the past eight years, IIT Madras strives to impact the entire nation. Through initiatives like the 10X program, we invite startups from across the country to leverage our facilities and resources. Our goal is to incubate at least 100 startups annually.

“I have been telling students repeatedly that 2025 on the first day of placement give me the pride of announcing that 20% of my graduating class are CTOs. This is the target that we need to work towards and I hope every institution aims at that as well.”

India can become a superpower only if we generate more employers rather than employees. India can become a superpower only if you own your ideas. So, patenting becomes extremely important. Innovation and entrepreneurship are the basic building blocks for a superpower BHARAT 2047.

Semiconductor manufacturing is a costly affair. How do you think varsities like IIT Madras can be instrumental to the policies like PLI scheme?

In semiconductor manufacturing, there are multiple steps. One is design, so there are a lot of design houses in India, setting up one is not that costly as the other steps. There are a lot of open-source tools as well. Also then comes the physical design. This process transforms the soft design into a Mask, using Electronic Design Automation (EDA) tools. This step often requires commercial tools and significant computational resources. IIT Madras has taken the initiative to establish infrastructure, funded internally, to provide cloud facilities and access to commercial tools. Startups incubated with us can utilize these resources to progress their designs.

Once the Mask is created, the fabrication and packaging phases follow. Packaging, which includes assembly, testing, and packaging, has seen significant growth in India. Many packaging companies operate within the country, covering these aspects extensively. However, the missing piece in India’s semiconductor ecosystem remains manufacturing itself.

Of course, we have the Semi-Conductor Laboratory (SCL) at Chandigarh. I am proud to share that a few months ago, we successfully taped out a chip from there, which is still functioning effectively. However, transitioning to a large-scale production facility operating at a high level of niche technology is a costly endeavor that our country is currently exploring.

Another important aspect is Intellectual Property (IP). While everyone owns a mobile phone, how often do we look at processor inside it? Peripheral IPs such as memory IP, Wi-Fi IP, and 5G IP production incur significant costs. If one intends to create a product, a chip alone isn’t sufficient; it requires a System on Chip (SOC). Essentially, there are numerous peripherals surrounding it, and all must be meticulously aligned. IPs are becoming increasingly expensive, and it is important for India to establish an IP program or negotiate deals where these IPs are sold with the promise of post-sales profit or royalties. Thus, there are two critical areas that IIT Madras is addressing, and the nation should consider when evaluating semiconductor manufacturing: fabrication and IPs.

IIT Madras received funding of INR 110 core from Sunil Wadwani to Architect AI and Data Science School. How does the institution plan to implement the funding in creating an ecosystem while enriching research enrollment and employability.

First and foremost, the AI and Data Science School started only three four months ago. We thank Sunil for his very gracious contribution. Actually, it is 105 crores that will be put as an endowment. This endowment is expected to yield 5 to 8% returns in the current market, with 5% allocated for departmental expenses. Annually, we anticipate returns of 5-10 crores from the market. With this funding, we aim to implement numerous innovative activities.

One very interesting activity we will be announcing shortly is a Bachelor’s Program – B.Tech Program in AI and DATA ANALYTICS. The admission will be made through JEE (Advanced). It will indeed be a world-class course ensuring world-class jobs. AI is not just computer science, AI is a mixture of multiple disciplines which is highly dependent on Data Science. Data has SYNTAX and SEMANTICS. SYNTAX is more of computer science and data structures whereas; SEMANTICS is much more on domain expertise. SEMANTICS needs a broad understanding of AI.  This course will have a world-class syllabus and will be launched for the first time by IIT Madras and I hope a lot of young students will be a part of it.

We are also starting many programs in the PhD research side. Specifically, we’re looking at responsible AI, which is going to be extremely important for policy generation etc. We’re also looking at many other aspects of AI in terms of policy, governance etc. It is one thing is to solve problems using AI and another to study the effects of AI on policies, governance and the environment.

AI for health, particularly AI for public health, is another area of focus for us. To just give you an example, we’re looking at domain-specific research in the area of AI. It is highly interdisciplinary, there will be cyber security component included. Currently, we are thrilled to announce that faculty members from seven departments have collaborated to establish the School of AI and Data Science. Sunil Wadhwani’s contribution has come at the right time, we’re very grateful to him. So the HoD of this particular school is the richest HoD today because he gets a yearly grant of INR 5 crore from this donation. This is truly phenomenal.