Tech Talk with Saumer Phukan (20+years at Intel) ??

Tech Talk with Saumer Phukan

(Director of Sales- Intel India Key Accounts)

I was reading up on developments from the Intel Innovation Conference where CEO Pat Gelsinger and CTO Greg Lavender announced the latest additions to OpenVINO , which is a free toolkit that allows developers to easily optimize, fine-tune and run comprehensive AI inference. Intel’s 13-Gen Intel Core?desktop processors also released recently, with enhanced features to support computer vision and AI capabilities. Enhancements were also announced for OneAPI - an open-source cross-architecture programming model that lets developers use a single code base across multiple architectures.

To learn more and delve deeper, I invited Saumer Phukan - Director of sales, India key accounts from Intel - to share his vision and mission towards growing Intel’s business in India in the space of data and AI.

Saumer is an experienced technology veteran who's been with Intel for almost 20 years managing diverse portfolios such as sales partner programs and vertical solution planning. Saumer drives the overall business consumption across India's enterprise, government, and cloud sectors.

?Some excerpts from the interview -?

Aishwarya: As you’ve said before, Intel is at an inflection point, can you share some of your insights on how Intel has changed and grown in the last 20 years??

Saumer: It has been a privilege and an absolutely unparalleled learning experience for close to two decades at intel across various assignments in multiple countries. Before I really go into what has changed, let's see what has not changed. We had the privilege of having an Internet pioneer, Vint Cerf, come for an internal discussion. And as Vint was talking to one of our executives here, he mentioned that once he heard “Software business is eating away at the Hardware Pie” to which, he responded, “Hey, the last I heard, all software runs on computing devices. So, the need for computers, the need for higher quality, more throughput computing has been more than ever before, alive and well.”

To this end, Intel continues to push the bleeding edge of technology to fulfill our mission, which is about improving the lives of every person on this planet.?

On what has changed, the problem complexity has increased. Because consumers have easier access to technology, their expectations on performance, cost, and security have really gone up. This puts pressure on organizations such as Intel to be unparalleled with our manufacturing leadership. And, we have doubled down on that.?

Moreover, due to the tighter interplay of hardware and software, we have this unique opportunity and are very focused on projects like Open.Intel - to make sure that we provide a complete stack, hardware plus software plus in some cases services to address the business problem of our customers versus earlier where we probably would have been primarily focused around hardware.?

The constant tussle between hardware supporting the software requirements and vice versa has always been a challenge and one which Intel always strives to solve.

Aishwarya: Can you talk a little bit about what are the key focus areas now for Intel and how is Intel using data and AI as part of its growth track?

Saumer: Now, if you look at Intel's history and journey, I can say that in a way, Intel liberated data. If you go to older data centers, each of them had a proprietary technology. What we did is we embraced Linux and that enabled “opening up” the data center for organizations and individuals to be able to access the data based on a standardized commoditized platform, which is what the industry is transitioning to.

The next critical requirement is Security. As you heard Pat (Patrick Gelsinger – CEO, Intel Corporation) mention, we are anchored on Open + Choice + Trust. Here, we have invested in making sure that we operate a zero-trust platform irrespective of whether the data is in process, whether it is in transit, or in store.?

If you really look at this then; Web 1.0, just about standardizing data, Web 2.0 was all about aggregating data. Web 3.0 is about preparing for a disaggregated distribution of this data so that organizations and individuals who create the data, also have the ability to secure, and drive the right kind of performance and monetization. Disaggregated distribution of data necessitates heterogeneity of compute. There is a major difference in the workload that you need for running intelligence on a camera versus the workload you really need to be able to run on a data center.?There are very different requirements to do that at a silicon platform level, we are beginning to create opportunities and hardware very aligned to what your workload is about.?

Now coming to Artificial Intelligence, one – which we all know – is one of the richest use-cases for data. Intel is treating this as a significant opportunity, and we are working to create a comprehensive bottoms up approach here.?

With the relevant heterogeneous hardware, the next conversation is about access. How do we expose the capabilities through an Open eco-system for the community to leverage, develop and maximize. As an example, one is our work with Google around the Open XLA project, talking about the low-level virtual machine compilers. That makes a fundamental difference on how you abstract the hardware capabilities versus the software applications that the developers and the community require. Second, what we also did with TensorFlow 2.9 release is we integrated our Open Neural Network platform - OneDNN into the Tensorflow build so that by default the optimizations reach the developers.??

In essence, step one, is the hardware. Step two, is access to the hardware capabilities through optimized the platforms and the frameworks like, for example, TensorFlow. The last piece is to build & deploy. You need to be able to build, and deploy your models with the latest data sets that you have. That's where you heard about platforms like iPDK infrastructure, processor development kit, OneAPI, and OpenVINO. These are frameworks that allow you to be able to build highly optimized AI models. These are some of the end-to-end capabilities Intel is developing in conjunction with the eco-system and critical to our growth and areas of investment

Aishwarya: I liked how you touched upon edge computing, distributed computing, how we're collecting data, how we're processing data, and how we are really using AI or ML as a tool to derive insights and make predictive models out of it. So, based on the use cases and technological requirements, what are the strategic partnerships that Intel has made??

Saumer: We have this unique opportunity as Intel right now to work across hardware plus software stack. Let's start at a component level. In India, we are very involved with organizations like CDAC for National Supercomputing Mission capability and similar other players to help them build up the platform.?

If you build on from here, at a service provider level, we are working actively with the Telco providers – Reliance Jio, Airtel, and Vodafone-Idea. Similarly, you may have heard about PhonePe, which is one of the largest digital payment providers in India.?PhonePe talks extensively about is the number of transactions and how much computation they need to be able to service customers and the exponential growth in payments happening in the country. If you combine this and the environmental condition in India, how do you drive data center design, how do you make sure you optimize your power cost, TCO is crucial. We are working with PhonePe on an innovative technology - Liquid Immersion Cooling that enables sustainable, high ROI data center design and operations.?

Further up the stack, at an application level, we also worked with one of the largest CRM providers, a unicorn hosted in this country. In CRM, one of the key capabilities you want to drive is to be able to use natural language processing (NLP) on text, to analyze the sentiment of the customer feedback and generate timely appropriate response. We worked with this customer to reduce their latency and improve throughput by ~4x.

Let's look at the public sector side as well, we have worked with top healthcare providers like Aster Healthcare & CARPL.ai to create a secure federated learning platform. What happens is it allows the data to reside closer to the creator and locally stored improving security.

Last but not least, I'll give an example of a solution for the transportation industry called iRASTE. It's an intelligent load and safety solution for highways in collaboration with Mahindra and IIIT Hyderabad. The intention being to use technology to improve road safety, to be able to cause up to 50% decline in the number of accidents that happen on the roads in Nagpur.

Aishwarya: What are Intel’s goals for up-skilling professionals to build and leverage these new industry innovations?

Saumer: Intel has a program called ‘AI for Youth’, a very well-established and recognized program in the Indian ecosystem, catering to about 2 million Indian citizens. And we've trained about 350,000 students on how they can develop AI based models and usages to solve the critical needs of the country.?

We're now expanding this a little bit more. We are working with the ministry of electronics and information technology (MeitY), to work on “Responsible AI” - a program for senior grade students, so they can learn how to use AI for the broader growth of the country.?

The questions on working with the ecosystem to empower lives of millions of people are somethings which are really personal & that I'm very passionate about! Thank you, Aishwarya, for asking the question.

Aishwarya: Visiting back to what you mentioned about the skill gap. Do you feel that there were any other roadblocks or challenges that you faced during these collaborations??

Saumer: There are three areas to which I would like to draw the attention to; first - Governance. When you're talking about AI and you're trying to accelerate capabilities based on AI for the broader market, governance would be key. Working with the government and public sector, the private sector needs to come together to make sure we're driving the right governance so that we are not compromising on security, sovereignty, and what's expected from a citizen.?

The second point I want to highlight is the conversation of skilling. I touched upon ‘AI for youth’, but if you really dig-deep, where we want to put India out there with the rest of the world, we actually have to get ahead of what the other markets offer. From what is required versus what is available in the market today, there's a significant gap. The skill-gap is another roadblock that I think we need to work on.?

Last but not least, building a strong ecosystem for component manufacturing and innovation. The government has done a great job of starting with initiatives. These incentives, combined with the training facilities, and building processes to drive quality, will help us build a brand that's dominant for generations to come.?

Thank you for the discussion.

Overall, it was a great experience talking with a senior leader to understand their vision and strategies for the growth of a major organization like Intel. For more on this topic, please check out the full video on my youtube channel ???