Startup Opportunities in the AI Revolution

by | Dec 7, 2023 | Video

A deep-dive into the AI revolution and the vast startup opportunities it will present.

This deep-dive conversation is with two AI Data experts, Vikram Joshi (Founder & CTO of Compute.AI) and Piyush Sharma (Serial Entrepreneur & Investor). They discuss the AI revolution and the vast startup opportunities it will present with John Furrier and Dave Vellante from theCUBE.

John Furrier: Okay, welcome back everyone. It’s Supercloud 4 here in our Palo Alto Studios. I’m John Furrier here with Dave Vellante.

We’ve got two great guests and one of the focuses we have here is founders and how to start companies. We have two experienced founders, Piyush Sharma, who’s the investor and entrepreneur, formerly sold his company to Tenable, just recently left there and is kind of freelancing out in the open here. I’m sure he’s got some grand plans, doing a lot of research. We’ve had him on before. And Vikram Joshi, CTO, Founder, and President of compute.AI. Guys, both CUBE alumni’s.

This topic here is important to me, because being an entrepreneur, Dave and I have seen the waves as you guys have. You’re legends, experienced, but also you’re out in the arena, and you know, the cloud was great. You put your credit card down, you can start a company. “The SaaS Revolution” was born.

AI’s different. We’re seeing stories of gimme two engineers and a product person, and one go to market. The recipe is faster. Is it faster? What do I use? How do I build the tech? What’s available? Let’s demystify this as founders. Welcome to this panel.

Piyush, we’ll start with you. As you look at the opportunities clearly, like the web, every vertical’s open for business. Right. What’s your perspective?

Piyush Sharma: Thank you for having me here, John first. Thank you Dave. I think this is one of the most revolutionary times we have seen in, at least in my lifetime, after the internet revolution. We are seeing significant changes happening on how the cloud have been used, especially now. The AI LLM, is a buzzword name.

It’s become super easy to build a company, as long as you understand what problem you are solving. There are sufficient amount of technology that is available, and then you can use new technology to build more technology. Like the AI LLM today is run from cloud, for cloud, essentially changing the game of the startups.

John: And you’ve been an investor, so you know the playbook. You put some seed capital in, you get some VC.

Dynamics are changing big time. What’s your perspective on the funding? How do you get funding?

Piyush: I think the market needs have significantly changed in the last few months and years. The investors always look at the white spaces. The bigger spaces like AI or LLM are similar to that. The usage of the whole AI ecosystem is going to be very significant. So, money-wise, I personally feel there’s plenty of money available. You just need to have a problem that is important and urgent, and that’s about it.

Dave Vellante: That’s the key. You still have to solve problems, right? And so, you’re in the world today.

So, what’s the process that you use as a founder to go about trying to understand a problem? I mean, a lot of times, we just fall back on what we know, but then you have to step back. I mean, you’re in the world of data. How do you go about figuring out what problem to solve next?

Vikram Joshi: That’s a great question, Dave. I think we are playing much like what Piyush said, at a very important time in the evolution of data processing, computing, AI, ML. And we are playing at a very important juncture here, at a confluence of two kinds of forces that. I just want to set the tone for a little bit of the background, and then we can discuss the option in what it takes to be an entrepreneur and raise capital in this….

John: Yes, set the table.

Vikram: …seemingly tough market. So, it’s the confluence of two major things happening here. One is you have massive amounts of data processing going on, and then there is the AI/ML forces, whether it’s GenAI, LLMs, or other kinds of AI/ML. And there are many opportunities here, and I know the market has been saturated with conversations on GenAI and LLMs. But there’s a much larger, there’s a vast amount of opportunity in the AI/ML space outside of that too.

From an entrepreneurial perspective, understanding what the problem space is, understanding the real pain points, how do we identify those? The architectural tenets of entrepreneurship haven’t changed. It comes down to team. We’ve talked about that. People have talked about that, there’s ample information. As an entrepreneur, you think you focus on the team, great teams build great products. And of course, you must have a market. And it’s not very easy to see the path in terms of how big is the market? Can I tap into it? Is it greenfield? Is it brownfield? What the opportunity is, and that’s where they cannot-

John: Did you raise money for your startup, or did you fund it with your own cash initially? Did you raise funds? And how did you capitalize that startup?

Vikram: So the urgency that I felt….. I did, I did, the first thing I did was put my own money in.

John: You put your own money in.

Vikram: I put my own money in, yeah. So I put my own cash in. I went ahead and got a million-dollar line of credit on top of it to make sure that whatever the market conditions, we have the ability to continue to focus on the idea and just go after it with gusto. Because once you believe, once you commit, you do not hesitate. And I think that that’s one of the first principles of entrepreneurship, right? You have to have skin in the game. It’s just between you, and it’s me and me, right? It’s not even how investors look at you. And then the next thing was, there were other investors who were interested, including people who invested in me before. So that’s how-

John: We talked about this. We had coffee prior to coming on the program, before, with some of our friends. And we talked about opportunity recognition, and that you don’t really need a lot of capital to go get going.

Vikram: That’s right.

Even the seed game has changed. You can either fund it yourself, because it’s like the cloud days, you put your credit card down, and you put it in the cloud. A little bit higher scale with AI, higher velocity. What’s the power dynamic on the funding piece these days? And I know you’ve been doing a lot of mentoring of other entrepreneurs as well. What’s the current state-of-the-art mindset and execution playbook for, you know, three feet in a cloud of dust, get some momentum, how I get that next check?

Piyush: Yeah, and I think, John, you made a very good point. Sometimes, it’s all about driving a value and innovation. I think when you deal with the new technology, or the new hype cycle of a brand-new technology like AI, LLM, there are investors, there are a sufficient number of challenges that there are up there to be solved by somebody, by you or somebody else, right?

I think the funding is a secondary aspect of a problem that you’re trying to solve. So, if you really start with the right problem, most fundamental, I keep on saying that, “Is the problem important and also urgent?” Is where you actually start. As an entrepreneur, when I look at anything to invest, if I am initially investing in a startup, I always look at that, the problem, that how important it is. Is this going to be for 10 customers facing the same problem every day? Or is it going to be 1000 customers, that sees this problem once in a month?

Vikram: And the urgency.

Piyush: Yeah,the urgency.

Vikram: Sorry to interrupt.

Piyush: No, absolutely. The important and urgent is where you need to start.

From the investment standpoint, just to wrap it up on the investment part, the investment, whether you bring in your own money, whether you bring in investors’ money, always consider it’s your own money, because that’s the conviction that the founder needs to bring in. Once you have that conviction, money will follow, the customers will follow, and everything will start fall in place.

Dave: Let’s talk a little bit about markets.

Vikram: Yeah.

Dave: You guys both compete in very large markets. You’re in the data world, you’re in the security world.

Piyush: That’s right.

Dave: But the data world is dominated by the big three cloud players, and then you get Snowflake and Databricks, which have seemingly reached escape velocity. So, the market’s enormous and it’s some big, big beast.

Vikram: That’s aptly worded, yeah. That’s the lay of the land there.

Dave: And in security, it’s highly fragmented. Really nobody has a double-digit share. Maybe Microsoft does, but generally speaking, it’s very, very fragmented. So completely different, you know? But both huge. So how did you go about thinking about the problem that you wanted to solve? You burned the boats, you’re not afraid of the big guys. Was it because AI was a potential disruptor, or you saw that the existing guard was not capable of doing what your vision sort of led you to?

Vikram: Yeah, I think just the framing of the question itself is very powerful. And understanding who the big players are, they’re the cloud players, they’re the big behemoths. There’s Snowflake and there’s Databricks. As an entrepreneur, I’m always looking to create value. And I don’t view disruption in, you know, in technology as being something that you decide upon. If something is going to change the game, let the customers vote on it.

So how did I make the decision out there? You are playing in a very rich data landscape where it’s a trillion-dollar market, or whatever the numbers are. It’s a massive, massive market. And no one denies the fact that there is plenty of opportunity. How do you go and insert yourself as a fledgling founder with a tiny little team, with tiny resources, and create value?

And in my case, I was thinking in terms of incremental, but significant value where you leverage the great work done by the Snowflakes of the world, by the Databricks of the world as, and you’re standing upon the shoulders of giants and creating value, which really means finding those pain points – real pain points like Piyush said, that are urgent and you want to be able to solve them.

And what we heard through all the conversations, and having been in the data-space, and knowing that I’m not going to come in here and build, you know, another computer engine. You don’t want to be the 10th search engine, and not do something like Google. So, at the back of your mind, you know that if the problem’s large enough, personally, it’s not exciting to me as an entrepreneur. So, you always have that push and pull out there. Is this large enough? And how am I going to insert myself? And is there going to be credibility? You come out there from the left field and you say, okay, here’s a Databricks, here’s a Snowflake, data management is done, you know? What’s the next thing? What’s the pain? And how do you go out and create value where anyone who buys your product probably already has one or both of these solutions?

Dave: Yeah.

Vikram: And you still want to out go out there and make it.

Dave: That’s clearly the case in your work.

Vikram: Security.

Piyush: Yeah, security is exactly like that. And honestly, on the security side right now, there’s so much noise around certain things, because everybody wants to do everything. But nobody focuses on the main thing, which is the prevention against the breaches. And every new technology brings an opportunity in the cybersecurity, right? AI came in, I mean AI/ML, ML has been there for a long time.

Dave: Yeah.

Piyush: The large language model, the performance and the efficacy improves in the last few months, like when the transformer architecture came in.

John: I want to ask you on that. One of the things Dave and I were just talking about before you guys came on was that CrowdStrike event.

Piyush: Right.

John: And all these events we go to at theCUBE. Everyone has the same line; I want to get your reaction to it….. “We’ve been doing AI for years!” (Piyush and John laughing)

Piyush: Exactly.

John: Okay, so they have-

Vikram: I have a story there too.

John: All right. (all laughing)

John: Yeah, you said the 10th search engine.

Dave: We just bought a company that’s doing that!

John: And your point about the 10th search engine is, Google was the nuanced point, but that’s like, they came in late to the game and became dominant. They had a different approach. What is the disruption vector that startups can take? If you’re fresh starting a company, you got LLMs, you got the foundation models.

Piyush: Right.

John: What about generative AI makes it a disruptive enabler? And how do you change the game and flip the script? Is it the data? Is it the app? Is it an AI wrapper? What does AI native mean? Because I look at all the old guard stuff and say, okay, CrowdStrike might’ve been there, done that, they’re doing it-

Piyush: Right.

John: Some say they’re the best at it. But is it really going to be the next wave? What’s your guys’ perspective of this disruptive opportunity for the startups? Flip the script.

Piyush: Yeah, great question John. Actually, you put it very well, that everybody says that we have AI, “We’ve already been doing AI for 20 years.” The challenge is the whole, why this disruption is happening now, because of the consumerization of the AI. AI and ML has been a very niche, very specialized platform in the last few years. But the whole launch of OpenAI, ChatGPT, has consumerized AI in the hands of actual end users.

Vikram: Very well stated.

Piyush: Right? The whole, so AI has become a commodity. A large language foundation model like that, there are like 10 of those foundation models available. And even to the extent that there are the models which are region-specific. Like there was one company got funded in Europe, who solves only the Europe-specific, the Europe-specific problems around foundation model. So, the whole commoditization, which is where everybody gets a hand on something or other, an application that uses AI, or being secured by AI, I think that’s the change we’re talking about.

Dave: It always starts in consumer though. Even in data, it started with the Google file system.

Piyush: Right.

Dave: In the search.

Piyush: Right. Yeah.

Vikram: Absolutely.

Vikram: And an example of that is, you know, I had someone who worked with me in the AI/ML space, is the lead engineer. Remember one of these applications in Gmail, as you type, it completes the sentences on you? That’s a massive generated AI LLM model using millions of whatever TPU’s out there at Google. And it’s been around for a long time. But it’s that commercialized digitalization and the commoditization. It’s like search and ChatGPT are now starting to compete. And I think that gives it wings.

John: So, I want to get back to-

Vikram: ML has been there.

John: I want to get back to the consumerization point because LLMs, and we’ve just showed our power law chart on our intro segment, that the size of the models are changing. You’ve got smaller models. To flip this script, I’m hearing the word proprietary and walled gardens a lot. And OpenAI was, is called the proprietary model. Actually, it’s more open. And other models in the long tail are proprietary intellectual property. So, data’s now the IP, okay?

Vikram: Yes.

John: So, you look at that. So, people are putting a walled garden around it, but yet it interacts with other data. So, the question is, how do you look at the data aspect of it from the perspective of security? Now, does security means securing against breaches, or just knowing that something’s truthful? What if synthetic data gets in there? How do you know what’s real and what’s not real if the consumerization is here? And to your point, my daughter was texting me a new app she’s using to write her emails for. She’s just graduated college, so they’re already using it. Apps are coming out, AI wrapper apps.

Piyush: Yep, yep.

John: What’s the next generation? What is that security posture? Is it about truth? Is it about breaches? All the above?

Piyush: Yeah, so I’ll take this, I’ll take a ten second attempt to explain that. What is an impact of this LLM on cybersecurity? And this is just my definition, there’s no standard on this.

There are four dimensions on which that AI LLM has been impacting cybersecurity now.

One is your cyber defense, which is where a lot of AI will be used to breach the system, creating DDoS attacks, or a lot of threats. Vectors will generate come out of the AI LLM. There’s a foundation model that created in the dark web. It just creates the malwares. So, you can imagine now, more and more data is translating to that.

Number two is cyber defense. So, a lot of AI usage is starting to happen to prevent breaches. Very contradictory, right? There’s one that’s used to attack, one is to prevent

And then there is a socially responsible aspect of AI that hey, whether by AI because of generative needs to be socially responsible, et cetera.

And the last is, the fourth dimension is preventing attacks on AI itself. So, if every company, if the planet is going to use an AI infrastructure, you need controls that will prevent breaches on AI.

To the last point, which John mentioned, the data. Data is so critical, because data has become the threat factor in the industry for the first time. It’s not the user, it’s not any, it’s not the infrastructure, it’s the data. So, you use data to exfiltrate the data. You create prompts so that you can exfiltrate the private, you know, the PII information. So that’s why it’s a very transient, it’s a transient time, but it’s a very significant change, the way the security solution has been.

John: And by the way, everything you just said is reality happening now.

Piyush: Yeah.

John: And the old standards were siloed data warehouses, old data modeling, old compliance.

Piyush: Yeah.

John: So governance, compliance all have to adjust.

Piyush: Adjust.

John: Rapidly.

Piyushi: Exactly.

John: And some companies are conservative, they don’t know what to do.

Piyush: Yep.

John: Yet it’s changing so fast.

Piyush: That’s right.

John: So the question is, what side of the street will you be and landing on? The good side, or the side that goes away?

Piyush: No, I think you have to lean on both sides, because we’re unfortunately or fortunately remain in hybrid mode for our longer period of time, like private to public data center. And I’m seeing AI as a complete trajectory as cloud. Initially there was challenges in adoption. There were risk chasms, which companies were built around solving the adoption problem. The same thing happening in AI also.

Dave: Yeah, well, as Vikram said, it all comes down to value. And the promise of AI is, it’s going to drive productivity up. I think in your world, the SOC analyst experience is going to, is changing-

Piyush: Absolutely.

Dave: ..dramatically. And then with so much more we can do with data. What do you guys think about the sort of timing of when we’re actually going to see? Maybe it’s happening already, but it’s still not meaningful in numbers. That we’re actually going to see AI have the impact that is promised?

Vikram: I feel that where we are in the industry with data management, these were massive problems creating systems of records, right? You can call them silos or not, does not matter. But solving those problems, the kinds of problems that say, Snowflake has solved, I think that’s been a massive step that we need to leverage and catapult ourselves to the next level. What happens beyond that?

And there is another point of inflection here, and again, you need to bring in AI/ML into the picture. It’s important. I do have a separate comment on the synthetic, versus the real part of the data. And that has to do with what’s happening with cloud data warehouses, with lake houses, data formats. The moment you take a data warehouse, or you take a table and you flatten it and turn that into, say, a Parquet file, at that point, you’ve unshackled compute from the confines of say, a database or a data warehouse. At that point, it’s, “may the best Compute engine win”. And there are many out there. But the ability to operate compute…

Dave: Maybe for users.

John: Well, yeah. Democratization.

Vikram: Right.

Dave: You don’t have to use the cloud vendor’s Compute necessarily you mean, right? That’s locked. And even though they’re separate, we’ve talked about that.

Vikram: And Compute for all right? Compute is the oxygen for data processing, right? I mean, we don’t pay pay for the air we breathe, and we are being productive, we are being entrepreneurs. We are creating incredible value, and hoping to change the world.

So, to have to pay for oxygen, I don’t think, you want to make Compute abundant. And we now have arrived upon the opportunity here where between Parquet and formats such as Iceberg, or Delta, Hudi, doesn’t matter. It’s data and metadata.

You can now have some very powerful compute platforms, which are far simpler, because I, you know, maybe this may not be that relevant, but at least I’ll sow the seed here for maybe a future conversation. What’s important here is, especially coming here with my roots in databases, having had something to do with Exodata. When I look at a database, you look at databases as having to do ETL is done, then you take the data, and you push that into rows, columns, tuples, whatever, into tables, right, things with schemas. There’s DDL and DML, there’s transactions, and then you have DQL, which is your query.

With Parquet and Iceberg, or Delta, which are very powerful standards, what you’ve done is you’ve taken DDL and DML out of the equation. You’ve taken transaction processing out of the picture, you’ve taken snapshots, checkpoints, everything out of the picture. And now you’re left with a pure read-only compute engine with the power of being able to do transaction processing. We are able to do point in time rollbacks, you know, all of those features of databases.

So, I think we have arrived upon a new feature. If I may just connect the dots on the AI/ML side of it.

Dave: Well, we got to wrap up, but I want to put a pin in that, we’ll follow up. “The Compute is oxygen,” I love that line. We have one quick question to end the segment for you guys as founders.

Vikram: For sure.

John: Other entrepreneurs out there, we are in historic time. It feels like the web, it’s going to grow. What opportunity recognition techniques should the younger generation watching this, or anyone in enterprise trying to figure out how to be less conservative, more aggressive without failing or getting over their skis, as they say? What advice would you give from an opportunity recognition? What’s real? What’s a company, not a feature? What would you guys say to that question?

Piyush: If I would start, I would give only one advice: talk to customers.

Talk to the real users of the AI infrastructure. That is how you will define how things need to be built. Talk to them. Talk as much as you can.

John: Right.

Vikram: And I would add to that, supplement that, and say talk to investors too, because they have talked to many, many founders. They’ve gone through 40 pitches or more through that a week. And what they possess is that information which you can rapidly access. I’d rather I’d benefit from the intuition of maybe five bright minds than 10,000 rational thinkers. And I think between talking to real customers and talking to, you know, smart strategic investors, you have it covered.

John: It’s a great point. And remember the DotCom bubble, not that it did burst, but it ended up happening. Everything that they’d invested in, it did happen. The spray and pray approach was the best. Spread the seeds out on the soil, let ’em grow. Just the amounts might differ.

Vikram: Absolutely.

John: Piyush, Vikram, thank you for coming on. I wish we had more time, great panel. Great to see you guys. As always, expert contribution here on Supercloud 4. Generative AI, generating tons of opportunities for entrepreneurs and innovators to change the game. The script will flip. This next generation is here and growing.

By the next interview coming Brian Harris, CTO of SaaS, a company that’s been around for a long time. Still private, does billions of dollars. His approach and his vision on how to take conservative approach, but yet aggressive with AI is up next.