play_arrow

keyboard_arrow_right

skip_previous play_arrow skip_next
00:00 00:00
playlist_play chevron_left
volume_up
chevron_left
  • Home
  • keyboard_arrow_right Featured
  • keyboard_arrow_right Validation
  • keyboard_arrow_rightPodcasts
  • keyboard_arrow_right Experimenting with AI in Pharma: A Question of Time [Vasyl Chumachenko]
play_arrow

Featured

Experimenting with AI in Pharma: A Question of Time [Vasyl Chumachenko]

Vasyl Chumachenko June 26, 2024


Background
share close

Yan Kugel⁠ is joined by Vasyl Chumachenko, Head of AI Consulting at Data Science UA. They delve into the future of pharma and explore the intersection of AI, ML, and GMP compliance. With a wealth of experience integrating AI solutions into chemicals and pharmaceuticals, Vasil brings a unique perspective on driving innovation and growth in the industry. Together, they explore the prospects, risk strategies, and practical implications.

Delving into the Future of Pharma

In this episode, we will delve into the future of pharma and explore the intersection of AI, ML, and GMP compliance. Today, we are thrilled to have Vasyl Chumachenko, head of AI consulting at Data Science UA, join us as our esteemed guest. With a wealth of experience in integrating AI solutions into chemical and pharmaceuticals, Vasil brings a unique perspective on driving innovation and growth in the industry. So stay tuned as we discuss prospects, risk strategies, and practical implications.

Vasyl’s Background and Expertise

Vasyl Chumachenko has a background in sciences and AI, with a PhD from Kiev National University. His expertise lies in integrating AI solutions into chemical and pharmaceutical processes, with a focus on improving efficiency and accuracy. As the head of AI consulting at Data Science UA, Vasyl has successfully accomplished projects in the domain of AI and chemistry, making him a valuable resource in the industry.

The Types of AI Systems Used in Pharma

When it comes to AI in pharma, there are two main subdomains:

  • General AI: This includes language models like GPT and computer vision models, which can be used for tasks such as automating processes and employee education.
  • Domain-Specific AI: This involves using state-of-the-art algorithms to extract value from chemistry, biology, and health-related data. This includes applications in drug discovery, where AI can help identify optimal compounds for further investigation.

Challenges and Opportunities in Implementing AI in Pharma

While AI has the potential to revolutionize the pharmaceutical industry, there are challenges related to regulations and validation:

  • Regulatory Risks: The pharmaceutical industry faces challenges in implementing AI due to strict regulations and the lack of clear definitions for AI in the pharma sector.
  • Validation: While validation of AI models is not necessarily difficult, the lack of clear regulations makes it challenging to implement AI for everyday quality control.

The Potential of AI in Pharma

Despite the challenges, AI has the potential to transform various aspects of the pharmaceutical industry, including:

  • Chatbots for Employee Education: AI-powered chatbots can facilitate employee education and qualification improvement, making knowledge more accessible.
  • Machine Learning Models for R&D: AI can be used to develop models for peak integration and chromatographic peak prediction, enhancing R&D processes.
  • Predictive Maintenance: AI tools can assist in predictive maintenance for chemical supplies, improving efficiency and reducing downtime.

Advice for Implementing AI in Pharma

o prepare the ground for implementing AI in pharma, companies should consider the following:

  • Data Management: Establish a robust data flow pipeline and store data locally or on the cloud for accessibility and security.
  • Collaboration with Service Providers: Provide expertise and advice to service providers to ensure the successful implementation of AI solutions.
  • Regulatory Compliance: Stay informed about regulatory developments and be prepared to adapt to changes in regulations related to AI in pharma.

Conclusion

The future of pharma is intertwined with AI and machine learning, offering opportunities for innovation and efficiency. While there are challenges in implementing AI in the pharmaceutical industry, the potential benefits are significant. By staying informed, collaborating with experts, and preparing the groundwork, pharma companies can harness the power of AI to drive growth and innovation.

Episode Chapters:

  • Introduction: 0:00 – 1:25
  • Introduction of Vasyl Chumachenko: 1:26 – 2:40
  • Discussion on AI and ML in pharma: 2:41 – 14:12
  • Challenges and advice for flawless implementation: 14:13 – 21:34
  • Risks and mitigations in AI implementation: 21:35 – 27:50
  • Impact on operators and need for education: 27:51 – 32:54
  • Conclusion and closing remarks: 32:55 – 34:13

Podcast transcript:

Please be advised that this is an AI generated transcript and may contain errors.

00:27 – 01:03
Yan Kugel: In this episode, we will delve into the future of pharma and explore the intersection of AI, ML and GMP compliance. Today, we are thrilled to have Vasyl Chumachenko, head of AI consulting at Data Science UA, join us as our esteemed guest. With a wealth of experience in integrating AI solutions into chemical and pharmaceuticals, Vasil brings a unique perspective on driving innovation and growth in the industry. So stay tuned as we discuss prospects, risk strategies and practical implications. So Vasil, welcome to the podcast.

01:04 – 01:08
Vasyl Chumachenko: Hello, Yan. Thank you for inviting me.

01:08 – 01:50
Yan Kugel: That’s my pleasure. That’s my pleasure. So, AI is a very hot topic at the moment. And a lot of people are highly interested to know what’s going on in the market. And you have a very unique position where you work in a company that works with a lot of clients, helping them establish their AI in their manufacturing facilities and different projects. So I think it would be quite interesting to talk to you about different aspects of AI in pharma and the diversifications that you see. And before we dive into it, Vasile, so you have a background

01:50 – 02:07
Yan Kugel: in sciences and also in AI. So how did you come to the position of head of AI at data science from maybe a bit overview of your background and what are you doing as head of AI data science?

02:08 – 03:06
Vasyl Chumachenko: Yeah, sure. So it’s actually a pretty maybe common story in our region. First of all, I got my PhD from Kiev National University, where I work on the targeted drug delivery, and it was mostly in nanoscience, and part of my PhD did this, what the algorithm based on convolutional neural networks where I applied this neural network to improve the very common method in nanoscience, namely dynamic light scattering. It’s also pretty applicable in some cases in pharmaceutical industry or basically the purpose of this method is to establish the particle size distribution of the dispersed systems. So sometimes

03:07 – 04:09
Vasyl Chumachenko: it is applicable in Pharma. And after that, I understood that What I really love about chemistry and the life science it is the parts that That calculate something so I really I really love numbers so I started to Dive deeper into the artificial intelligence. And that’s how I switched from the academic position to AI outsourcing business. And I started as a consultant. I mostly help data science UA with my domain expertise in chemistry. And I successfully accomplished a couple of projects in this domain. And that’s why I’m here now as the head of AI consulting team.

04:09 – 04:15
Vasyl Chumachenko: And I really love artificial intelligence though, basically that’s why I’m here.

04:16 – 04:32
Yan Kugel: Great. So AI has a lot of applications and there are also different types of AI. So can you elaborate on the most common AI systems types that are used today in the market? Used today in the market?

04:34 – 05:34
Vasyl Chumachenko: Okay, so we could divide the artificial intelligence in pharma or in chemistry, let’s say in general, or in the industry into 2 subdomains. The first 1 is a general domain where we have our beautiful and extremely useful child GPT or another language models. And we can talk about computer vision models for different purposes in pharma. It is in pharma and in industry, actually, in general. That is well known for public that we sometimes can use GPT like models to automate some tasks for the employers and it’s like general tasks which could be really, really useful for

05:35 – 06:34
Vasyl Chumachenko: any business And on the other hand, we have a really specific, really domain specific artificial intelligence and machine learning models where we apply a real estate of the art algorithms to extract the value from our chemistry related or biology related data or health related data. It is actually pretty obscure domain, so basically no 1 except the professionals know about these algorithms, and we can talk about both. So, first of all, the large language models, like ChaiGPT or Gmini, they could be really, really useful in pharma, where we are talking about, we can use these models to

06:34 – 07:41
Vasyl Chumachenko: build a perfect system for the employer’s education, like we are able to fine tune the large language models and into this specific knowledge database. So this data would be easily accessible through the chatbot interface or whatever the customer want. That’s on the 1 hand, because everyone knows about chat GPT and the basic capabilities, but it’s not so common that this model can be fit to exact the data set to have the opportunity to use this data set precisely, but easily for the, for example, for the employers. So it is actually, I believe, a beautiful application for

07:41 – 09:00
Vasyl Chumachenko: the large language models on the 1 hand. But we also could talk specifically about the special purpose machine learning models. So maybe we could speak about different domains starting from the drug discovery process where we can actually use sophisticated so-called graph neural networks to extract the, basically to find the optimal heat compounds for further investigations for the in vitro trials, for preclinical trials. But the first stage we have to choose the most potentially biologically efficient molecules. And you know, nature allows to have more than billions of billions of different organic compounds. And it’s actually a huge

09:00 – 09:19
Vasyl Chumachenko: challenge to choose the optimal molecules for the certain purpose, again the certain disease, again the certain target. And the artificial intelligence is currently helping pharmaceutical industries to choose the best options.

09:20 – 09:46
Yan Kugel: So you have experience in this domain of research, especially in the AI, as you explained, a lot of research and discovery of a molecule. So do you also implement AI at the moment with different clients in the domain of manufacturing, QC, quality assurance? So is it something that is also quite needed at the moment and popular?

09:48 – 10:55
Vasyl Chumachenko: It is on the 1 hand and it is not. Here I believe we should address all the risks starting from the regulatory risks and basically every other risk because it’s all about the real people and final consumers, final drug consumers. In quality control, We know that a lot of equipment manufacturers that manufacture the basic equipment like chromatographers and other analytical equipment for pharma, for example Agilent or Waters, they already incorporate the artificial intelligence and machine learning algorithms into their software. And on the 1 hand, this is a really good software. They have their really powerful algorithms

10:56 – 12:02
Vasyl Chumachenko: for, for example, automated peak integration or for the chromatogram prediction for the certain mixture of the compounds. On the other hand, we could check out the FDA regulations on the use of the artificial intelligence and quality control in general. And here we could easily understand that neither FDA nor European regulator have even a good definition of the artificial intelligence in terms of pharmaceutical industry. That is why it’s like the pharma industry are in a peculiar situation when you have to decide if you even could correctly use this software for the quality control purpose or in fact,

12:03 – 13:00
Vasyl Chumachenko: they are mostly are able to use it only for the R&D purpose. For example, when you have to develop, when you have to design the best possible analytical method for the, for example, to measure the exact concentration of the API. And in this case, these models, they became really risky for the manufacturers and for the majority of the situations as I am aware of and I have expertise working in the project like that as a part of Data Science UA. We see that that manufacturers, they really want these models to work on the back in R&D

13:01 – 13:53
Vasyl Chumachenko: but not in a real quality control. So only for the preliminary stage, not like on everyday basis. That is why it’s really risky for manufacturers to apply it. It’s like regulations are strict. These regulations came from early 2000s and the new regulations, they are basically not ready yet. So yes, we’re Talking about quality control, these models are really state of the art. They are powerful but they are still risky for the manufacturers and that’s why they are not really in demand.

13:55 – 14:10
Yan Kugel: So you would say they’re risky because the regulations don’t address those types of software yet or also because you feel they are not precise enough and very difficult to validate them?

14:12 – 14:36
Vasyl Chumachenko: Talking about this model validations in our opinion like as a company, as a service provider, the validation is not really so difficult. It’s not so hard to implement. The problem is that regulations are really not ready for this technology.

14:38 – 14:43
Yan Kugel: And do you see change happening at the moment on the side of the regulators?

14:44 – 15:44
Vasyl Chumachenko: Yes, yes, sure. For the last couple of months, I read the summary of the huge of the huge conversation inside the European regulator and they came to some conclusions of what has to be really improved to really allow pharma to use this powerful technology on an everyday basis. But still, even after this conversation, it’s like, okay, let me be honest, if the challenge is to define the artificial intelligence and machine learning for pharma, it means that we really have a problem because we maybe have to wait 1 more year until the quality control process could be

15:44 – 15:47
Vasyl Chumachenko: really widely automated by machine learning.

15:49 – 16:17
Yan Kugel: Right. So do you think that companies right now are testing all the features on the trial basis and getting ready to implement them also in full manufacturing? And do you think that companies that will not do it will stay behind because they will be not able to be quick enough in establishing those technologies?

16:20 – 17:29
Vasyl Chumachenko: I believe that the biggest innovative companies, they not only test the existing software. I strongly believe that they are developing their own solutions that would be perfect for their purpose. And for this, I am sure they have really good reasons. The first 1, it’s always better to keep the data in the house if you have an opportunity like that. On the other hand, another part of these discussions between the different stakeholders, for example, European pharmaceuticals, these so-called low impactful models like for R&D, they won’t be really hard, they won’t be regulated as hard as the models

17:29 – 17:58
Vasyl Chumachenko: that would have an extreme impact on the potential patients’ health, like the artificial intelligence for the clinical trials, which also could be a pretty powerful tool, but also the most risky tool. So yes, I believe the big companies are into these machine learning algorithms. But the smaller companies, they probably would like to use some hand from the experienced outsourcing companies.

18:01 – 18:40
Yan Kugel: Right. And do you think that small companies can use the power of AI to surpass huge companies that might need a lot of time to establish systems globally? And we know that AI can assist in cutting hours of man work time, right? So it can be a huge advantage. So do you think small companies should take advantage of this possibility and just explode in inefficiency by leveraging those systems and playing around with the possibilities?

18:42 – 19:38
Vasyl Chumachenko: I believe it is a possibility. First of all, because smaller companies have an eternal bureaucracy, which is not so hard to bypass by the internal stakeholders like the R&D laboratories or something like that. And they are more open to the collaboration with IT company, with the service providers. And because they mostly would like not to invest in having the internal specialists like machine learning engineers or data scientists. And that is why I believe that it is a possibility. But still everything depends on how hard the real regulations on this part of the machine learning technologies would

19:38 – 19:45
Vasyl Chumachenko: be in future because currently they are on the first stages, I mean the regulations.

19:47 – 20:34
Yan Kugel: Right. So I know that many software companies try to service the pharma. They work on different AI tools to assist with the manufacturing and automate a lot of processes. For example, equipment maintenance, prediction of risks, prediction of deviations, etc. So how familiar are you with such different solutions And do you think there are such solutions that can be more easily implemented because they’re less control the production but more give information which is helpful and does say require less strict validation, for example?

20:36 – 20:42
Vasyl Chumachenko: I believe you are talking about predictive maintenance purpose of the-

20:42 – 20:43
Yan Kugel: For example.

20:44 – 21:47
Vasyl Chumachenko: Yeah, I mean, for this exact case, yes. But it really depends on the, again, it depends on the exact purpose. Because if we, because from time to time, API synthesis in industrial scale, scale could be still pretty dangerous for the environment, pretty risky for the employers. And In this case, when the health and even life of the workers depend on some sort of artificial intelligence, it could be risky. On the other hand, this exact technology could potentially make the life of these workers better. So, yes, if it is less regulated. But I believe we cannot generalize

21:47 – 21:50
Vasyl Chumachenko: here because it really depends on the exact process.

21:52 – 22:06
Yan Kugel: Okay. And do you have examples or use cases where you worked on projects or heard of projects where different AI tools were implemented in different GMP activities?

22:09 – 22:15
Vasyl Chumachenko: In GMP activities, I mean, you mean in predictive maintenance, for example, yes.

22:16 – 22:26
Yan Kugel: Quality manufacturing, anything that is being used during the manufacturing of a commercial drug?

22:29 – 23:33
Vasyl Chumachenko: Only for the chemical supplies. Chemical supplies. When the synthesis of some chemical supplies are still under GMP, but it is not exactly the pharmaceuticals. In this case, these suppliers, they are open for this technology and they are ready even to collaborate with service providers. And yes, that is possible for some chemical supplies for predictive maintenance of chemical supplies or for quality control of chemical supplies. Yes we do have and and still it was a really challenge task for us as for service providers. Mainly because we have to bypass this hell. But yeah, let me be honest,

23:34 – 23:58
Vasyl Chumachenko: It was a hell because we were waiting a lot of months for the security, internal security to allow us to start working with their data. Okay, so it was hard, but it is possible. Let’s be honest here. Yeah, it is a hard task.

24:00 – 24:31
Yan Kugel: So, yes, I’m sure that they’re working with such regulated industry. It’s hard, especially the big companies. So what types of uses do you see? What use cases can you tell us about of not only maintenance for predictive maintenance, for example, but any kinds of AI that you see being used in pharma, chemical, QC? What can we do with this, if you can give different examples?

24:32 – 25:41
Vasyl Chumachenko: Yeah, starting from the AI-empowered chatbots that facilitate the eternal education and the qualification improvement of this app, where we use basically 2 technologies, large language model and the so-called RIG, retrieval augmented generation. So we can fit today large language models to the knowledge data set. This is a real use case and we helped with developing it because we have to start small. But also We developed 2 models, 2 machine learning models for the R&D department, which mostly work with the chromatographic data, the first 1. We developed a model that performed the peak integration of HPLC data

25:41 – 26:35
Vasyl Chumachenko: really accurately for the exact purpose of the company. And another for the same company, we developed a model that pretty accurately predicts the chromatographic peaks of the compound mixture depend on the different solvents. So yes, we do it, but again, it is still a challenge to implement it on the quality control, but not for the R&D. Because for R&D, it’s okay. For everyday quality control, it is hard to tell.

26:37 – 27:20
Yan Kugel: Right, so based on the successful projects that you finished despite the challenges, Could you give some advice on how to prepare the ground to make it as flawless as possible? So let’s say a company wants to implement a model and they hire a company, a consultancy, that service provider that helps them write the code and implement it in their system. So you mentioned there are a lot of hassles there like getting all the access to the tools, to the IT tools, et cetera, et cetera. So if you could map the challenges and maybe give advice on

27:20 – 27:24
Yan Kugel: what to keep in mind to make it as flawless as possible.

27:25 – 28:23
Vasyl Chumachenko: Oh, yeah, sure. I mean, it is still a challenge for pharmaceutical companies to establish the pipeline of their data flow. And they have to save every data point possible. For example, if a company records the infrared spectrum on an everyday basis, they have to save everything from the original files to metadata. And if they want to automate this process somehow because from time to time it is the case. They have to store everything. They have to store it either locally or like on the cloud and it is a completely another challenge because okay here what I

28:23 – 29:28
Vasyl Chumachenko: remember how the okay couple of days ago I spoke to the pharmaceutical community of Ukraine on the conference and 1 of the speakers, actually my colleague who represented the IT community, he talked to the big Ukrainian pharma go cloud. And actually pharmaceutical industry should move their data from local storages to cloud because it is both safer, faster, and it is basic advice. But I believe it works for every company which works with the big data. On the other hand, no 1 except farmer professionals really know what they want from the artificial intelligence or any other tool.

29:30 – 29:49
Vasyl Chumachenko: If a farmer works with service providers, they have to be ready to provide as much advice and expertise from their side as possible. It’s a perfect background for the successful project.

29:52 – 30:15
Yan Kugel: Right, so they need to be ready to open the files and give the full access to the service provider to make it as flawless as possible. And this is some issues that you see quite a lot that companies give you hard time to get access to different tools and networks.

30:17 – 31:09
Vasyl Chumachenko: I mean that is still the challenge. It is a challenge for us because we have to be really careful about every bit of data. That’s why I told that pharma have to go cloud or if they have a possibility to set up a really powerful local storage if it is the necessity. Because Yes, we would need to have an access to the data on the 1 hand, but access to the data and the possibility to, God forbid, share this data with illegally. I would like not to have even a small possibility. That is why we should

31:09 – 31:58
Vasyl Chumachenko: have the possibility to have the access to the data, but only the possibility to access, not the possibility to share. On the 1 hand. On the other hand, of course, we will guarantee the data security. But still, it is a challenge. Because on my best knowledge, even a pretty big pharmaceutical company, they don’t really have the perfect data infrastructure. And it is a challenge for them and for us We have to prepare Prepare the data for the let’s say industry for 4.0 because it is the next step. And I believe in 5 to 10 years, a

31:58 – 32:11
Vasyl Chumachenko: lot of processes will be automated by machine learning, artificial intelligence and pharma. Maybe in 20, but still it’s, you know, 20 years, it’s not a lot really.

32:11 – 32:55
Yan Kugel: Right. So from what you’re saying, when a company chooses a service provider to help them establish new systems and develop AI, they need all the access, but they also need to make sure they choose the right partners to work with because we’re talking about security and there are a lot of risks with that. So how would you recommend pharma companies, chemical API manufacturers to choose their partners and qualify them in the best possible way. So what criteria would you recommend them to look at when they’re choosing such a partner?

32:57 – 33:57
Vasyl Chumachenko: The couple of criteria, but some of them are pretty obvious. It’s a general reputation of the company, but I know that it sounds like too general. So basically, they should look the first if they basically really have an experience working in this domain because it is really complex. But I understand that not every outsourcing company has this experience. In this case, they need to provide an assistant for this company. But okay, another 1, another point, If it’s like, maybe we should call about the red flags, red flags here, here, because if company, service provider really doesn’t

33:57 – 34:15
Vasyl Chumachenko: want to, to speak, to tell about every possible risk of implementing these models on each step. That means that you should run away from this company because I afraid that this would be potentially really risky.

34:17 – 35:01
Yan Kugel: Right, So have a partner that is clear about what they can do, what they cannot do and that the company doesn’t afraid to lay out their risks and don’t just paint everything pink because we’re talking about activities that can be risky and time consuming and also complicated. So this is a challenge because you train the AI to do some tasks and it takes time with the data and you want to make sure it’s efficient at the end, right? And if you already choose 1 partner and they don’t work out, then you lose a lot of time,

35:01 – 35:05
Yan Kugel: a lot of money, right? So it’s a very important part.

35:05 – 36:00
Vasyl Chumachenko: In some cases, yes, but another good point would be here is that from time to time, we should start from the small projects. And it is really possible, especially when we are talking not about the GMP activities, but if we want to automate some internal processes In pharma, I’m talking about management, about like the auto filling CRM systems, etc. It is, of course, it is strongly regulated internally, but it’s not so sensitive. And from time to time, we could start with a small project which could be delivered in a couple of months. And after that, we

36:00 – 36:05
Vasyl Chumachenko: could go further. So start small because it is always the case.

36:06 – 36:50
Yan Kugel: Right, makes sense. So test the companies, see how it works out, see their expertise, see how well you’re doing with the people, how well is the communication going. So that’s a very good advice indeed. Start small, then develop to bigger projects. And you mentioned there are risks and some companies don’t completely put them on the table, don’t reveal everything. So from your experience, what are the most common risks that accompanied and they want to develop such models for the QC, for example, what should they take into account when they do their risk analysis and so on

36:51 – 37:01
Yan Kugel: because they might not know what AI is exactly. That’s why they’re talking to you, for example. So what are the risks they need to consider?

37:02 – 38:03
Vasyl Chumachenko: The first risk is the data available for the company because again, as you told on the beginning of our conversations, that validation of the model is a really, of the artificial intelligence or machine learning model is a really complex task. And to validate the model, we have, we need to have a perfect data to validate it on. And we need to have it really diverse. We have, we need to have a lot of different, like well-balanced data that represents the real world, let’s say, as close as possible. And the main risk is that we don’t really

38:03 – 38:56
Vasyl Chumachenko: have enough data. And on the other hand, if we are talking about another machine learning applications from another domains, Sometimes we don’t really care that we took a part of data from the public sources, but for the pharmaceuticals, especially for the quality control, it might be not the case. So we should rely only on the data provided by the client. And the main risk is that we probably don’t really have enough data and good enough data to really build the perfect machine learning solution. That is really the main risk. Another risk is of course, that the

38:56 – 39:42
Vasyl Chumachenko: artificial intelligence, especially narrow networks, These models are not really interpretable in the sense as any another model. In other words, we don’t really know why the algorithm really gives 1 or another decision. At this risk could be mitigated in the only 1 way that we have to use concept human in the loop. Because every step of the artificial intelligence have to be controlled by the human.

39:44 – 39:55
Yan Kugel: Right, so Basically, although you build a machine to be as accurate as possible, you still need a human to check and confirm the results. So that’s what you mean.

39:56 – 40:33
Vasyl Chumachenko: Exactly. That is a mandatory stuff. It also the part of the absolutely necessary part of all the discussions inside the entire artificial intelligence community in the pharmaceutical industry, both organized by FDA or European regulators. So yes, we still need a human and we will need a human and human beings will not be substituted by the machine in this industry for a really long time. And you know, it’s a good news.

40:34 – 40:52
Yan Kugel: Right. And so when you have the output, so does the output look always the same as it looks right now? Or would the operators would need to learn new skills to understand whether the results are correct?

40:54 – 40:54
Vasyl Chumachenko: I

40:56 – 40:57
Speaker 1: mean,

40:59 – 41:46
Vasyl Chumachenko: you would need an additional education for the employers, but it’s always the case when we are talking about the new technology. On the other hand, for example, if we are talking about automated peak integration in chromatography, it would be not a big problem because the automated integration exists for a really long time. And for most of the employers, even well-educated with the deep knowledges, The previous models are also work boxes as well as the current machine learning models. So in these terms, no, we don’t really have an extremely complex process of the effort to get educations

41:49 – 42:12
Vasyl Chumachenko: for the employers. Of education for the employers. But yes, in general, people need to be educated in this domain, in the domain of artificial intelligence in general and in the domain of these specific technologies. So, because they need to understand what is possible and what is not.

42:14 – 42:43
Yan Kugel: Okay, Good. So in regard to a pharma regulation in GMP, so you said that there are a lot of risks still and companies don’t go willingly at the moment to start developing different tools because the regulations are not clear yet. So where do you see developing the next couple of years? So when do you think there will be a breakthrough in this regard?

42:44 – 43:37
Vasyl Chumachenko: In this regard, I believe that, okay, if we are talking about pharmaceuticals in general, I believe that the breakthrough would be in the domain of drug discovery, because the first stages of the drug discovery, they are not really so regulated as in other stages. And that’s where I believe we could… I personally waiting for the breakthrough, maybe it’s already here, we just don’t know yet. But I know even some Ukrainian companies, they are like chemical suppliers for pharmaceuticals, they are already using this algorithm. They partner with powerful product developers in this domain. And that is really

43:37 – 43:54
Vasyl Chumachenko: promising for me. So I expect that I personally waiting for this breakthrough. I believe it lies here, but in another domains, I believe we would have a really not the revolution but step-by-step evolution. So it

43:54 – 44:06
Yan Kugel: will take more time testing and new regulations testing and it probably will take more time right in the more regulated manufacturing domain, right?

44:06 – 44:18
Vasyl Chumachenko: Yeah, I believe that currently manufacturers, they are testing a lot, but they would use it widely in a couple of years.

44:19 – 44:59
Yan Kugel: Okay, so that’s quite good. But I think, anyways, I think all those companies should be on their toes. So they should not sleep while the regulators think they should test and also already be ready once regulations are there because the breakthrough that AI can do in those domains is massive, right? So there should be testing it and already being ready to implement it because at the end the regulations will come and to understand how to implement it through risk management and with oversight of human. As you mentioned, there is a lot of AI that you can

45:00 – 45:28
Yan Kugel: implement in parallel to humans to process data, right? And as long as you have a human signing off all the things at the end, it should be okay, right? So I think that should be the direction at the moment. And I think we are in a good path there. So, Vasyl, do you have any points that you think that you would like to add that we didn’t discuss yet to this interesting conversation?

45:29 – 46:07
Vasyl Chumachenko: Well, I mean, we already discussed the main points I wanted to mention from the very beginning. But I would like really to encourage pharmaceuticals, starting from the small biotech companies and to giants to experiment and where it is possible, where regulations allow it because it is just a question of time when the machine learning will be everywhere in this industry and nobody wants to be aside from this process.

46:07 – 46:20
Yan Kugel: 100% Vasily. And if somebody has questions, would like to reach you, what is the best way to get hold of you or any 1 of your colleagues?

46:22 – 46:33
Vasyl Chumachenko: The best way is to check out my LinkedIn page or write me an email. Of course, it could be shared.

46:33 – 47:03
Yan Kugel: Right. So you will find the information in the description of this podcast. So you will be able to connect with Wasil, who is a very good expert on AI and maybe chat with him and follow him also to see the content he posts and about the updates in the AI domain. So Vasyl, thank you very much for coming to the show and talking to me about the different projects you’re doing in AI and the insights in the industry.

47:04 – 47:08
Vasyl Chumachenko: Yeah. Thank you for inviting again and have a good day.

47:08 – 47:09
Yan Kugel: Thank you very much.

Tagged as: , .

Rate it
Previous episode
Post comments (0)

Leave a reply

Your email address will not be published. Required fields are marked *