Brandy (00:01)

Today we are thrilled to welcome Harini Gopalakrishnan, the global CTO of the Life Sciences Division of Snowflake to the podcast. With over 17 years of experience in life sciences, Harini has built a remarkable career at the intersection of technology and healthcare, delivering mission critical results through innovation, product strategies, program management and implementation.

Harini's passion lies in leveraging cutting edge technology to solve complex business challenges in healthcare and her journey spans the globe, managing diverse teams across geographies and functions. Before Snowflake, she made an impactful contribution at Databricks, Palit... Okay, okay, okay, it's okay.

Harini (00:38)

Thanks.

I didn't, I didn't. yes, sorry. You might have to redo that part. No, no,

no. I was at Sanofi and I had managed vendors like Databricks and Palantir while at Sanofi.

Brandy (00:59)

Okay. All right. You know what? Okay. I'm going to just take that part out. We're going to just start this again.

Harini (01:10)

Wait,

did I send you a profile or was it my mistake?

Brandy (01:15)

You did not send me a profile. pulled this from LinkedIn. so if there's anything else, so the only, that was it. Was there any other inaccuracies before I jumped into that? Okay.

Harini (01:19)

Okay, got it.

No, no, no, that was fine. So

baby, what I can do is, can I share you?

Brandy (01:31)

Yes, of course. If you put it in the chat.

Harini (01:35)

Yes, I can just put that in the chat. Sorry, probably should have sent that before because

Brandy (01:38)

No, no, no, it's fine.

I should have asked for it. I just was trying to keep things off of your plate, as best that I could.

Harini (01:45)

I'm not able

to, I'm not able to send, you know what I will just reply to your email.

Brandy (01:50)

That's great. And I'll pull up my email.

Harini (01:52)

because

it's not letting me paste.

Brandy (01:55)

No problem.

Harini (01:56)

I just sent a tone to you.

Brandy (01:58)

Excellent.

It's loading. It's taking its time. Here we go.

Harini (02:16)

the email, yeah, always.

Brandy (02:27)

Okay.

Great.

Okay, excellent. I...

Take this and replace that.

So before I get into this, it's.

Palantir, is that right? Palantir?

Harini (03:12)

Yes, that's the vendor that I had Databricks in Palo Alto, I think it should say that I managed vendors like Databricks.

Brandy (03:15)

Palad here.

Okay, then cognizant is the, okay.

Harini (03:22)

Yes, and

the pharma and the company that I worked for was Sanofi, the pharmaceutical. Yeah. I sent the whole abstract, I hope I didn't just cut it into half.

Brandy (03:27)

Santa Fe. Okay.

no, you're good. have previously, she was in an industry advisory role with AWS.

Harini (03:37)

Yeah, in with AWS and then exactly Sanofi is the company where I worked

with Palantir and Databricks that was while I was at Sanofi.

Brandy (03:46)

Santa Fe. Is it Santa Fe or Santa Fi?

Harini (03:47)

Yeah, different.

It is Sanofi. You say it's Sanofi. It's a French company, so Sanofi.

Brandy (04:00)

Okay, great.

to start this again. Bear with me.

Harini (04:05)

No worries.

Brandy (04:12)

Today we are thrilled to welcome Harini Gobalakrishnan, the global CTO of the Life Sciences Division at Snowflake to the podcast. With over 17 years of experience in life sciences, Harini has built a remarkable career at the intersection of technology and healthcare, delivering mission critical results through innovative product strategies, program management and implementation.

Harini's passion lies in leveraging cutting edge technology to solve complex business challenges in healthcare. And her journey spans the globe, managing diverse teams across geographies and functions. Previously, she was in an industry advisory role with AWS, led real world evidence at Sanofi and headed up research informatics practice for Cognizant. She has worked with big data technologies within the life sciences domain and has leveraged

Planeteer, AWS, and Databricks from this perspective. Her industry expertise includes genomics, protomatics, protomykes.

Harini (05:18)

You can ignore that, you can see what you're doing.

Brandy (05:24)

Her industry expertise includes genomics, protomics, lab informatics, clinical, and real world experience and worked on solution. Sorry, sorry, it didn't add that on here. All right.

Harini (05:33)

Evidence. Evidence. Evidence. Evidence. Yeah. No worries.

No worries.

Brandy (05:45)

She has a master's in bioinformatics and an engineering and computer science. She is passionate about driving innovation in her domain, especially with Gen. AI and LLM like protein folding. Her AI based RWE platform in Sanify won the 2020 Gartner's Ion Innovation Award. Harini, welcome to the podcast.

Harini (05:45)

thing that.

Thanks for having me, Brandy, and nice to be here.

Brandy (06:11)

Yeah, excellent. You've had a remarkable career. It was actually really amazing to just dig into this a little bit, even just preparing for your introduction. Can you talk through a little bit about your journey and how it has shaped your approach to innovation in life sciences?

Harini (06:32)

Sure, first of all, thanks. think my career is more of not staggering, but it's something more of learning on the job all the time and every time you seek for something new. So, but looking back, it seems like a longer journey than what I feel as I sit right now. However, so personally, when I started, was a tech computer science major, an undergrad like most of us growing up in that generation in India where, but...

The interesting thing is that when you take computers back then, data scientist wasn't a fancy word, but I was looking to more use it for solving a specific business problem. And rather than developing code for a technical solution or a product. So that's how I got into doing my masters with my informatics. so that was a great learning. So I think that's the beginning of my passion or love towards the industry itself, because studying how

proteins worked in the body or how genes operated was a big learning because I wasn't a biology major in high school. So that was a completely new world and it just was marvelous. It seemed like a miracle that we can live and breathe. Every single thing has to work in order or work precisely the way it should work for us to even make a single, like for me to talk or for me to even have a walk or see or anything. Everything seemed like a precise clockwork machine.

Brandy (07:34)

Yeah.

Harini (07:55)

Using IT or data crunching back then without cloud was interesting. So a lot of the innovation that I saw back then are now larger size because of the fact that we can do things on a cloud. But I always feel and I always say that in every conversation that life sciences itself, my nature of it is transformative. That people have always employed, even if it's not fancy AI or machine learning, even if

you call it statistics or modeling or simulation. People have always used that for long. They used to do it on-prem in compute clusters. they, with cloud, it brought easy scale and ability to do it larger size, as I said, with ease. So for me, it's just an evolution. And that's why I love this industry and being in this industry, working closer to the use cases that are innovative has also helped me understand the technology that are innovative because you always need a

the next big thing out there to solve some of these complex problems because the data that you deal with is diverse, it's images, it's text, it's numbers, the compute that you run on them was huge. always, you always can find a need for a very innovative new technology to bring value. So that's why I've been able to follow that through and every job I do is always, or the next move I make is always about learning and what can...

I learned from that job that I'm actually looking for something that's a bigger role, for example.

Brandy (09:28)

Yeah, no, absolutely. think it's a really interesting, know, coming from a data background, but then bringing this into the life science world and just having that understanding from kind of the data lens into life sciences is really interesting. And I know that you also have been on both the customer and vendor side, and that's a rather unique insight that not a lot of folks have.

Can you kind of talk through that and how that has guided your role as CTO of Snowflakes Life Science Division?

Harini (10:03)

Sure. it's always nice to be, I mean, I was a consultant before I became a customer. So was heading up, as you said, an introduction or small research informatics practice. So I have consulted for most of the big pharma for their research informatics needs. So Sanofi, which I later joined, was also one of my customers, know, days I was heading up their practice. when you're a consultant, it's always what I believe makes a good consultant is being able to think

for the customer or be in the customer's shoes or, you know, be able to articulate or show, demonstrate in the meeting that you understand them very well, right? Communication is important, listening and making them feel that you know what your problems are, what your pain points are, I mean, just repeating what they said in a better way or in a way that is solvable or understandable goes a long way. So that...

I think is innately needed for every good consultant. always already think of problem solving and you already think on behalf of the customer. But when I joined the customer, what it got me was the appreciation for the constraints in which they operated. Because you think as a consultant, this is a problem, this is a solution. Why is it not obvious that you just go implement the solution? And when you're a customer, on the other side, you actually understand some of the constraints, organizational constraints, budget constraints.

maybe bureaucracy that happens in every big, large corporation. So things that may prevent you from implementing the best possible solution that you as a consultant would have thought it was a no-brainer. So that part, understanding the dynamics of a big all, being able to navigate that, building stakeholders, networks and collaboration, working with the right vendor. When you're a customer, you're also

you can see through the BS filter. So you already have, you always have vendors pitching to you and then you already know because you've been there done that you actually know who might sound legit and who might not and who's not got all the eyes dotted and these cross. So, you know, asking those questions. So those are some of the things that really helped me. And, but what I really learned was that how do you navigate the organization? Right? How do you make sure that you're even if the solution is great, the problem is very innovative.

Brandy (11:59)

He

Harini (12:23)

Is that what the company needs at that point in time? If that's what the company, if that value prop is not coming clear, how do you make that business ROI visible? Right? What data points do you collect to convince management? Managing up is extremely important. Working closely with your business is extremely important. So I did not know real world evidence coming into Sanofi. I was coming from the research informatics world. learning that, right? Sitting close to business and learning that domain and

seeing how technology can help the domain, always instilling the fact that the business is the king. We are working for them to solve their problem to make it easier. It's not like technology is not the king in pharma. It's an enabler. So having that mindset, managing up stakeholder constraints, those were the things I learned. So when I'm now on the vendor side, which is a great transition because that was the next logical step. You've been a customer, you've been a consultant. Now you've got to be a tech vendor, right? It helps because every day when I have sales and they go sell, it's always good to remind them that maybe we think

This is the best solution. I'm sure the customer thinks that's the best solution, but are they right stakeholders to make the decision? Do they have the budget to make the decision? Are we helping them in demonstrating the value of the platform that we are trying to bring in to their business counterparts for them to be supporting in their decision? So these things you only can, you don't think of it logically as a technology vendor because for you, this is the best technology. It's a no brainer. Other people are using it. Why is this customer?

not seeing it, right? But the advice that expertise I can bring in is to think for them in their shoes and say, hey, these are the constraints they might have. So let's help them, our champions be able to tell a better story and let's help them give that messaging in a way that they can build confidence internally. So most of my job is about that change management aspect and telling that story to an industry lens.

Brandy (14:14)

Yeah, it's a, it's a really interesting perspective that you have Harini. I'd say that there's, your experience is very rare. It's really hard to find people that can see all the different angles and, to kind of not, I mean, it really sounds to me as if you're,

Harini (14:21)

Well, thank you. I hope so.

Brandy (14:35)

You're not necessarily looking at one thing as being the only, right? You're taking all the various stakeholders into consideration and all the different factors and, and figuring out what you're not blinded by all the other things that oftentimes blind people when they're in certain roles.

Harini (14:53)

I hope so. mean, it could also be that I'm overthinking the problem. So I mean, I have to be aware that sometimes you don't have to solve for everything and maybe it works. But yeah, I like the fact that I can think through the different lenses. And that's probably what I think would be the unique skill that helps me. if I want to create a brand, that's what I would like it to be that you can see from different lens for sure.

Brandy (15:17)

Yeah. And something that you brought up earlier that I think is really interesting is you mentioned that life sciences, they've always used AI long before it became a buzzword. Can you elaborate on that a little bit and talk through how it's been applied and how it will be applied today and beyond?

Harini (15:38)

So basically, Lifesense is a many different sub-industries by itself. At the highest level, I've had to break it into three, which most people do, which is research and development, R &D, which is all about finding the drug and testing it on animals and then on humans and making if it works and prove that it's not toxic, you make a submission from the regulatory agencies to launch it to the market for everybody to use.

And then there is manufacturing, which is about making manufacturing the right way, the exact precise modality of the drug. And then commercializing, which is to take it to the field and make sure the doctors have it for prescription or they are aware of it so they can prescribe and pharmacies have it, et cetera. as I said, at the outset, there's R &D, there's manufacturing, and there's commercial. And they all deal with different things. And most of the time in a big pharma, these three operate almost like different sub-organizations.

Brandy (16:34)

Yeah.

Harini (16:34)

there

is not a need for collaboration. It's changing. People will talk about enterprise data, enterprise strategy, but still the decisions are pretty much made independent of each other. The budgets are made separate. There might be data sharing, but many, most of the time they tend to operate within that function. So if you ask where AI has been used, I would say within R &D, there is a lot more room for innovation, especially because when you talk about research, it's less regulated.

because you're actually in an exploratory space, you're trying to see what might work for what combination. there are a million possibilities. There's a new target or a target is basically what you study for a disease, right? That gets researched on by academia. There is a publication, there are tons of publication coming out every year that studying a disease or saying that this might work, that might work. This might be an interesting protein to go after. So, so much of wealth of information that's coming and in research, you want to be able to search that.

space and then find something that's interesting and promising that you want to take or move forward so that it can become a drug later on. And there is also, every pharma is going after, oncology is something that every pharma is studying. So you also have to be there actively researching and experimenting so that you actually can find that secret sauce soon.

Brandy (17:59)

Hmm?

Harini (18:01)

And that's why there is more innovation because you want to throw every tool, every trick up there and see to make sure that you can go after it. You find something that is more promising than somebody else. And the faster you do and use AI or any other techniques up there, you have a better chance of success in clinical trials and later phase because it can get costlier. So for a drug to go to market, people say it's an average 16 years endeavor from the time you first study it to the first time you actually

Brandy (18:22)

Hmm?

Harini (18:29)

get it approved to manufacture it to sell it. It could be 16 years minimum. And the later you go, so sometimes a drug can fail five years later on. Sometimes it can fail 10 years on. you studied it in animals, it works. You go to the first human, it's toxic and you pull back, but that's already like eight years to 10 years. So the more later it fails, the more costly it is for the company. And because of the effort, the time it takes to have gotten there and the cost of doing it in humans is much higher.

So you want to be able to find drugs early on that have a better probability of success later on. And to do that, you have a lot of data points. As you said, like I have, I think through every lens. Similarly for a drug, you want to collect every data point. You want to test it on a cell of tissues, right? Much earlier to see mimic what might happen in a human later on. You might want to get with a newer diagnostic, with a newer technique, sequencing techniques, you want to study the biological

break or smaller components of it like omics or proteomics within that particular drug interaction with your cell much earlier on. Now you can also do a lot of image analysis where you can actually see how the drug moves through your body or what changes happen to a cell when you inject that drug much earlier on. Because you can do a lot of electron microscopy scans and analyze those images, fantastic things. So back even in...

90s and 80s. People have always studied these drugs using machine learning through on-prem clusters and compute to see how they might interact in a body using experimental data points through simulations. They call it in silico drug design, which means you're designing it not on a human, not on a lab. You're literally designing it synthetically in a computer. So there are companies like Schrodinger, it's very well known, that have helped automate these algorithms and packaged it up so people can run these

Brandy (20:19)

Hmm?

Harini (20:27)

modeling algorithms using back then Linux operating system, run these programs, submit these commands that runs on a cluster and get outputs that feed your decision making later on. So that's when machine learning or you can call it machine learning started. And now you just scale it up. Now you add more and more data. You can add genomics, can add proteomics, you can add images. You add more and more data points, all with the intent. You can use generative AI now even too.

create new synthetic versions of drugs or find how a protein might look using AI or predict how a protein might look using AI, things that you didn't have to do in the lab, things you had to do in the lab before. So it's just improved on creating more data points that you can actually use to make decision making. So AI is ubiquitous, but even back then, if you could model and write an algorithm to do some prediction, people did it and that's where it started in research.

Brandy (21:06)

Yeah.

Yeah, that's a pretty incredible journey. mean, that's the ability to be able to do that in the nineties. It seems like nearly impossible, but I get it. And that had been dedicated to that particular field from that time. And now we're just expanding upon that and the capabilities. I don't know if you're seeing a future of just more predictive use of

Harini (21:34)

But yum.

Brandy (21:53)

AI technology, guess, where are you kind of seeing this shift and kind of come into, what is this going to look like, I guess, for the potential future?

Harini (22:03)

So, AI has two aspects, right? There is predictive and now which everybody talks about which is generator. So the predictive AI always has a future. mean, most of the tasks are predictive. As I said, in research you could predict whether this compound will be active or not, small molecules of the drug. Back, as I said, early 2000s as well, you had algorithms that could do it. Now it's generator, which is you can also design new compounds or proteins based upon certain input.

So that's where it has changed. But in the other areas, which we never talked about, like manufacturing and commercial, the innovation wasn't that advanced as much as research, because research, there was no regulatory pressure. You could move fast. Time to insights was more important than, you know, the regulatory guardrails. But in manufacturing, for example, you could only move so fast because it was very important. There's a lot of...

Brandy (22:45)

Mm-hmm.

Harini (22:58)

constraints because of the fact it's called GMP, good manufacturing practices that say that you need to be able to reproduce every single batch the exact same way that is outlined in the protocol. Obviously you don't want anybody to distribute drugs that are not manufactured exactly the same way that you know is safe or effective for use, So you can only innovate so much there. Or if you innovate or if you try AI, because AI is always, remember, heuristic. It's an approximation. It's probabilistic. It's not an exact rule.

based inference. So you want to be sure that the prediction is absolutely certain that it's true before you put that into a process. So there is a little bit of innovation that can only do so much because you need to validate it so many times. But predictive algorithms are there as well because you can now kind of predict early on. For example, a good use case of manufacturing is what's a golden batch? Like how do you

look at certain data points coming from temperature sensors and say based on which combination, what is the ideal batch parameters are going to be? What would give you the best performing batch or how do you perform the yield to be optimized on which conditions, which staffing point conditions based on predictive analytics? In commercial, you want to use predictive analytics for figuring out which is the, should your field sales

Brandy (24:12)

Hmm?

Harini (24:24)

focus the efforts on. what is the, for a certain geography, you have a certain position, but they have a certain population that's aging or a certain, what do call social determinants of health that point to some kind of disease. Say they're dealing with obesity, for example, you might want your field sales there to focus on a drug that might handle obesity, or you might want your field force to focus on a drug that has diabetes, estimate. So looking at.

segmenting patient population or segmenting HCPs and providing them information on what is the next best action to do, whether to reach out to the HCP via email, whether via do a direct call or send them a PowerPoint or meet them at a conference, helping them guide on their decision-making or where predictive AI is used. And generative now has just helped in.

helping with content creation. So now for all of these things, want content to go and talk to some HCP. want in the manufacturing world, you want content to be generated for FAQs, for example, or you want to know what is the right SOP for a certain deviation. Things like that is where the generative aspect is used. So predictive, 100%. It's implemented already even with the manufacturing commercial. It's going to grow.

Generative is now the one that's bringing the operational efficiency, which is the next step. I've predicted this outcome. Now, help me write content. If content is one of the things, what do I do is help me generate the content.

Brandy (25:54)

Yeah. And you bring up a really good point too, just in AI kind of always being a thing and R &D, but lagging on the, on the manufacturing side, if we're kind of breaking it down into three parts. And I can see why, right? The manufacturing side is, it's very delicate. I mean, it's, it's very complex. mean, not that R &D is not complex, but when you're manufacturing at a mass scale,

and trying to create the same consistency. I can't even imagine what that, the complexity of that. So can you talk through how the manufacturing side is now adopting technology, AI technology, and kind of what that looks like in terms of consistency and kind of bringing things to scale?

Harini (26:47)

Yeah, so the manufacturing, first of all, when we say manufacturing, we're clubbing a lot of things in it. Now there are different modalities of drug itself. It started with pills, which we normally call small molecules. Then we have antibodies or things that we call large molecules, are things that your body makes themselves, proteins. And the third now is...

the cell therapy, which is essentially taking something from your body and modifying it so that it can go and, especially for cancer, it can go and find the ones that are not confirming like a cancerous cell and use your own body's mechanism to fight them by doing some tweaks with your own T cells. So it's changed. So the manufacturing process for a small molecule, which is probably like your Tylenol or pill that you take every day, is different.

from manufacturing something like a biologics or like a vaccine, for example, which is potentially a modified strain of the virus that you have to inject. And then the cell therapy is completely different because you're not manufacturing in batches. You're taking every patient sample and then working on it and then sending it back for ingestion. So the complexity arises from the fact that even if you're one big pharma, you just can't say that I'll have four sites and all the four sites, both.

I take care of all manufacturing there. All these whole plants will do it. It's not like that because depending on each pharma you've got to recognize they will be dealing with these different kinds of modalities. There'll some that they will definitely might have some vaccines. They will definitely also have the small molecules because they will always exist. They, small molecules have been there forever. So that's what they've gotten into the initial existence. So they have to continue manufacturing that.

which nowadays if it's over the counter, you spin that up as consumer health and most pharma try to spin that out. But for more other reasons they exist and they have to manufacture that. And then there is some large molecule. Most companies still invest in that, antibodies, monoclonal antibodies and things like that. And then everybody knows getting on the gene therapy bandwagon. So you at least have four modalities and sites that have to be, each plant has to be producing that. And each production will have to be so, so precise.

because these can potentially have life threatening impact and it's regulated. One deviation, you would have a site closure. You'll have a big ramifications. So you've got to make sure that the SOP, is a specific process. Each site and plant has to make sure that they follow the process precisely documented, recorded, validated. You could have an audit anytime where the related agencies come in and you've got to prove that your standards are impeccable and.

The only way to do that is prove that every single thing you do is recorded. You can submit evidences of information from your machines that synthesize it from the notebook lab notebooks that run these experiments. People who are working there are trained. So you've got to provide a ton of evidence to make sure that the data is reproducible and you have everything under control. So with that regulatory pressure, as I said, it's not easy to change things. Even a simple tech upgrade.

If you do it, you've to do through a validation process to prove that after the upgrade, everything that you touched upon hasn't changed. you run a certain experiment, you've got to prove that the output is the same. So that's a lot of effort, right? So change management is huge. So those complexities add to why manufacturing can't push forward a lot of innovation. But it's changing because there's a lot of companies that can do...

Brandy (30:24)

Hmm?

Harini (30:29)

streaming, can get data real time, the technologies have improved, but you can stream the data real time. You want to bring operational efficiency because there is more and more need for certain drug types. And you want to be able to scale up your manufacturing so that there is no drug shortage. The only way to do that is incorporate AI somewhere to bring up these efficiencies, productivity improvements, et cetera. So a lot of companies are investing in making sure that they can modernize these plants, especially the newer plants.

There is no problem because you can build it with that in mind. The challenge is going to be in the older plants where they're already operating with a certain SOP. How do you go back and improve, include these modernizations because there is a change in that.

Brandy (31:01)

Right.

Yeah, absolutely. And do you see companies like Katalyze, right? Kind of taking their approach to helping out that manufacturing process. Can you kind of talk through that, how organizations like Katalyze are assisting on the manufacturing side?

Harini (31:18)

Cough

So basically it would be where there is a need for vendors is that you need to drive these automation and repeatable automation. And most of the time it deals with reading unstructured data, being able to do lead optimization, being able to find the golden batch, for example, things like that. So companies that can actually invest in AI and reproduce reproducibility in a way that

they can implement it easily for each customer without them having to invest in rewriting the code. There is always a space for making that happen. And hopefully, we haven't seen Katalyze in action as in complete production version, yes. But the intent of Katalyze, I believe, is to make that AI easier for customers to implement. So there is a room need for that. And then the question is, how easy it is, how accurate is it, how reproducible it is, and how

how much they can scale to different modalities. We talked about a lot of that playing into its rollout.

Brandy (32:31)

Yeah, and thinking about Snowflake too, right? So how is Snowflake helping drive this innovation?

Harini (32:38)

So a lot of our customers deal with data because Snowflake at the very beginning is all about AI data cloud data analytics, bringing compute to data. So it all starts with data at end of the day. So the idea of Snowflake is like a performant analytics engine. So you can stream data, whether it's structured, semi-structured, unstructured. So the manufacturing that means sensor data, data from your...

MES systems or your LIM systems that are in the lab, that are in the plant. And being able to analyze them fast to make these decisions quick so that you can make post-corrections before you actually create the batch. Say the batch takes five hours to run and then at the end of time you figure that's not, that's a wasted one, you've got to speed on the batch again and that's delivery for the end pharmacies. So you want to catch some of these errors early on so you're not wasting time in.

creating a batch that's not going to work. So detecting that requires access to the data and being able to analyze the data really fast, right? And making those insights preventive. You want to be preventive, not reactive. So you want to make those analytics really fast. So that's where Snowflake helps is in, we have a separate manufacturing focus, which is both on process, which is what pharma deals with, but discrete, which is automobiles and others. But I can speak more only on the pharma side.

Brandy (33:56)

Yeah.

Harini (34:01)

So most of the conversations are around integrating with these systems, making sure the data lands in Snowflake faster, and then letting our customers get to the insights faster. And for that, there are native capabilities that Snowflake can do with AI, but we also work not just in manufacturing, but with an R &D commercial with partners that can bring the life science focus, bring some of those. For example, life science would have its own data type, like as we said, gene data or proteomics data. And so we need...

partners that can understand this data and do the transformation and run certain algorithms that are specific. So same thing with manufacturing. would like partners that can look at, say, a GMP documentation, read that, that, or do a golden batch analytics very specific to the features of interest for a pharma and do that. So also looking at partners that can integrate to make that happen quickly.

Brandy (34:51)

So that's where Snowflake and Katalyze partner. Yeah, yep, absolutely. As you're talking, I'm thinking about just the change and shift also just in people, right? The talent that is required for this shift in this industry. Can you share some perspectives there? Like, are you seeing more?

Harini (34:54)

That's one of the discussions you'll be having.

Brandy (35:17)

Data analysts, folks who have more of a holistic view and able to speak the different languages that kind of all three components, the different languages that the three components speak.

Harini (35:31)

So I think the change management is not as much as we thought it would be before because there's a lot of buzzword on AI, especially in life science. said that this year's Nobel Prize in chemistry was for a generative AI work for life science and research. it's literally something that's proven and people have embraced it. now, of course, as I said, lot talked a lot about within R &D, it's more easy to embrace in manufacturing. The pressures of

being very operationally efficient, the pressures of having to meet the growing demands of certain kinds of drugs in the market, there is a need to adapt so that there is no resistance. I think the point is where there is a lot of changes in terms of being able to articulate the business value and making sure the change can happen seamlessly. So there is a lot of

things that in, especially in the world of manufacturing, things have run a certain way for so many years. And now there is a need to change that. you, there's a need, even if it's not a big transformative change, there is some in-court patients you have to do, and that's a lot of work. At the same time, when you do that work, you have to keep things running. So managing that and how do you manage that skillfully? think that's the struggle that everybody has. Of course, as I said, if it's a new plant that you set up, it's easy to automate ground up.

Brandy (36:29)

Yeah.

Harini (36:53)

It's easy when you start something new to incorporate the latest and greatest. The challenge again with manufacturing is there's a lot of plants that exist and that are running. How do you go back and modernize that? So these are some of the conversations that are happening and it's not really a mindset change, it's more of how that is where the struggle is.

Brandy (37:12)

Yeah, and the plants that are already in existence, these plants are very expensive to build and to get fully operational. So it's not like an easy lift to start from scratch. So I would imagine that a lot of this work is going to be with legacy plants and manufacturing and then figuring out how to slowly make these updates and integrations.

Harini (37:33)

Exactly.

Yep. Exalted.

Brandy (37:38)

When you are thinking just about emerging trends and opportunities, can you kind of talk through just how you think that these will be shaping the future of the industry?

Harini (37:55)

I think we talked about it, we touched upon that a bit early on, but from an emerging trends, the manufacturing space, it's all about smart manufacturing. It's about automating the shop floor from ground up, being able to do a lot of preventive care, as I said, being able to figure out that some parameters are off, you might not get the golden batch and course correcting it, getting alerts before it happens. There's a lot of need, especially with cell therapy to...

build a cell therapy command center, which is because there is so many touch points involved in creating a cell therapy modality. You have the lab, have the plant itself, you have the person who's administrating the processed cell therapy back to the patient. So there's so many touch points and there's a shipment involved of going from the place where you collect the sample to the plant, the sending the finished shipment back.

lot of touch points that has to happen. And you need a place where you can monitor all of that, right? And then make sure that there is no break in the process somewhere that is making sure the samples you collect are good enough for you to actually do the lab work on and then do the manufacturing on and stuff like that. So there is a lot of need for technology building there and advancements there. So a lot of the direction I see is about integrating the data quickly, using AI to do preventive predictive maintenance.

Brandy (39:15)

Mm-hmm.

Harini (39:19)

Maybe some generated a way I could do authoring and productivity improvements, right? But, or being able to ask a question like an FAQ agent or something like that. But most of the work is going to be around automating the shop floor, making sure the data is available for analytics faster, using that for predictive calls so that you can make take preventive actions. And for a very complex modality like cell therapy, building a plane where we can get feeds from every single interaction.

and again make decisions so that you're not wasting time anywhere because of a information, failed capture.

Brandy (39:50)

Yeah, yeah, making a, yeah, the decision,

the faster decision-making, see this just being a huge, a huge push.

Harini (39:57)

Yeah.

That's the

primary driver for all faster decision-making to make safe trouble later on.

Brandy (40:07)

Yeah. Harini, as we kind of wrap up this conversation, can you just talk through the legacy that you hope to leave in your role at Snowflake since you've done...

Harini (40:15)

that's a big word. No, I wouldn't say like, you see, no,

I mean, it's, just like what I do. I like talking like this. like dealing with customers because I learned from everything I said. It's not something that I've been there. I've not been in a manufacturing plant myself, but I talk to customers. I learned from the interactions. I learned from the problems and I learned from vendors on how they solve the problem. So it's nice to be putting things together, right? I'm learning every day. So I love my job. So

Brandy (40:39)

Hmm?

Harini (40:42)

no legacy or anything of that sort. I just hope that I can keep learning and keep sharing that knowledge. That's all.

Brandy (40:49)

Well, thank you so much for sharing all of your knowledge with us today. And we look forward to having you on again down the road. Yeah.

Harini (40:55)

thank you. No worries.

It was nice talking to you, Brandy, and good luck to Katalyze as well.

Brandy (41:01)

Yeah, thank you, Harini.