Dr. Andrew Giessel is the Director of AI and Data Science at Moderna, where he oversees the team applying machine learning and statistical techniques to improve the company’s mRNA medicines, including the COVID-19 vaccine. Prior to Moderna, Andrew served as the Director of Data Science at Sense AI, a startup developing a platform for geospatial mobile sensor data. He holds a PhD in Neuroscience from Harvard Medical School, and a BS in Biochemistry & Computer Science from the University of Kansas. Andrew can be found on LinkedIn here and on Twitter @giessel.
Andrew’s expertise sits at the intersection of algorithms and the science they enable—two distinct purviews that actually work hand-in-hand. Why? While earning his neurobiology PhD at Harvard Medical School, Andrew noticed that his computer programming skills actually enabled him to better analyze the results of his research. In fact, he believes DNA is like code, and proteins are just small self-assembling machines.
Andrew’s two areas of expertise converged at Moderna, where he was the company’s first data scientist. Today, he spends time on computer-based analysis and modeling, while also working on experiments. Andrew notes that, even as Moderna has grown as a company and faces new problems everyday, the computational and quantitative approach of these two fields remains constant.
While most people today associate Moderna with the COVID-19 vaccine, it actually started as a biotech startup with AI at the heart of its drug discovery and manufacturing processes—with the essence of the company’s products stored in a computer as a string of letters and code. It was this inherently digital, AI-powered infrastructure that allowed Moderna to accelerate the research process and ultimately produce its COVID-19 vaccine so quickly.
How? To make messenger RNA (mRNA)—an information molecule that carries DNA instructions to your cells—Moderna buys small pieces of DNA from labs around the world and implements a quality control review to correctly combine them in its own lab. Moderna engineers have built two algorithms to optimize the process. One detects the optimal place to break apart a gene, so that they have the best chance of reassembling it into the correct sequence. The other correctly analyzes the poly-A tail of a gene, helping scientists more efficiently determine the quality of that tail.
Andrew likes to think of Moderna as, first and foremost, a platform with adaptable AI-powered solutions that have the potential to solve a problem over and over again for years to come. From the company’s work on flu vaccines, to the manufacturing of the COVID-19 vaccine, Moderna’s AI infrastructure enables the company to create a number of vaccines because the only difference between them is their mRNA sequence. With the equipment already in place, Moderna can simply change a mRNA text string at the beginning of the research process and iterate from there. Vaccines are a crucial application of Moderna’s technology, but there are so many others that come from the same roots: enzyme replacement, gene editing, oncology and more.
Moderna recognizes that, as the role of AI grows at the company, all employees should be educated on the basics of algorithms, statistical thinking, data collection and other critical aspects of machine learning. In fact, Moderna will soon be launching a first-of-its-kind AI academy through a partnership with Carnegie Mellon University that offers multi-day workshops. The courses will be built for employees at each level of the company, so that everyone will have the right language to talk about AI when an opportunity comes up—whether that’s in meetings with potential stakeholders or in strategic planning efforts for the future of the company.
Andrew is excited about the prospect of integrating AI into a variety of business processes at Moderna, including the expansion of its use in mRNA production. Once a doubted modality of therapy, mRNA has proven its efficacy in part thanks to AI, and is now being used by many startups to tackle problems in the medical industry. But Andrew also expects Moderna to incorporate AI into in vitro studies, clinical trial design, epidemiology studies, protein engineering and design, natural text document generation and more.
Across the medical industry more broadly, Andrew believes AI’s role will continue to evolve and expand. In fact, many startups are now leveraging the technology in radiology and pathology imaging to help doctors make better, more informed decisions. Andrew expects we’ll continue to see this partnership grow, as AI offers new angles on data or makes the tedious parts of their job less tedious—enabling them to see patterns or insights that the human eye would otherwise overlook.
EPISODE 33: Andrew Giessel
Jim Freeze Hi! And welcome. I’m Jim Freeze, and this is The ConversAItion, a podcast airing viewpoints on the impact of artificial intelligence on business and society.
On today’s episode, I’m thrilled to be speaking with Andrew Giessel, Director of AI and Data Science at Moderna. We’ve heard quite a bit about Moderna over the last year for its landmark achievement developing one of the first COVID-19 vaccines, and more recently for its ability to adapt the same vaccine to emerging variants.
Andrew has been with Moderna since 2016, and now leads the AI and data science team, applying machine learning and statistical techniques to improve the company’s mRNA research process—a feat that helped Moderna develop the vaccine in record time.
Today, Andrew will walk us through all of it—from how Moderna’s roots as a biotech startup helped accelerate vaccine development, to the growing impact of AI on the pharmaceutical industry. Andrew, welcome to The ConversAItion!
Andrew Geissel Thank you for having me.
Jim Freeze Yeah, we’re thrilled to have you. So you have a pretty unique background spanning both data science and biology, and now sit at the intersection of science and algorithms if I can say that at Moderna. Can you walk us through your background and what ultimately brought you to the company?
Andrew Geissel Yeah, sure. If you were to ask me as a child what I was going to do when I grew up, I would’ve said something with computers. It took me a little while to figure out that that was biology, but that’s the path I took. I started it as a computer science major at university in college. And a couple years into that actually I took an amazing biology course just to fulfill a requirement and the teacher was fantastic. And that’s where these big ideas of DNA as code or instructions and proteins as little self-assembling machines that do everything in the cell. Those big ideas clicked for me, as part of that class and I was just fascinated.
So at that point, I decided to get an additional degree in biochemistry, started doing undergrad research, where I used computers to simulate the motion of proteins and have been on that trajectory ever since using computers to study biology. I took a really amazing graduate-level seminar on ion channels, which are the proteins in neurons that allow them to have electrical activity. I was very struck by that and applied to PhD programs in neuroscience wanting to study ion channels specifically. And so I was accepted at the program in neurobiology at Harvard Medical School. And I ended up floating up higher in the sense of, instead of studying proteins, I studied synapsis, the connections between neurons, between cells.
During my PhD, and again, in my postdoc, I did experiments that simply required being able to program to analyze them. They were like electrical traces or movies with many images that changed over time, and we needed to be quantitative about how those movies or electrical traces changed over time. And so that’s where I think my training as a biologist and a computer scientist really started to come together. It was in my PhD when I was able to use those skills to analyze my experiments.
So after my post-doc, I joined a friend’s startup. It was a mobile phone tech startup, basically collecting and analyzing mobile phone sensor data. That startup didn’t last long. We ran out of money, and the next job I had was at Moderna. It’s been a pretty remarkable journey. I joined when there were around 300 employees. Now we’re over 2,500. We were at 700 people at the beginning of the pandemic. We went through an IPO. The company had an amazing trajectory and it’s been really cool to be a part of it.
Jim Freeze Yeah. It has an incredible trajectory and it’s obviously had an incredible impact on all of our lives. I’m just listening to you talk about your story. I love how you brought together two really distinct passions in a very interconnected way. As director of AI and data science, what’s an average day, if there is such a thing, look like for you?
Andrew Geissel Yeah. That’s a great question. It’s changed over the years. I was the first data scientist at the company. And when I started the job, I did a lot of analysis and modeling. My background as a computer scientist and as a software engineer, led me to help have a hand in some of the digital infrastructure we have. So it’s been a mix of thinking about experiments and writing code, and that’s moved to more managerial tasks over time. Spent a lot of time in Outlook and on video calls.
Jim Freeze Don’t we all?
Andrew Geissel Yeah. But the thing that has stayed the same is this approach of thinking quantitatively and computationally about problems. And as the company has grown, the types of problems that we face have grown as well. When I joined, we had just started our first clinical trials, which was a flu vaccine, H1N1 flu vaccine, I think. And so, there’s still a lot of open science questions. It’s not to say there aren’t open science questions still today. There’s tons, probably more than ever, but that was the primary focus.
But as we started doing more clinical trials and in the summer of 2020, when we went from a phase one trial through a phase three trial, and we had not done a phase three clinical trial at that point, the company just kept on growing, the opportunities and the data that we are collecting just kept amassing. And so my day-to-day job, it is a lot of meetings, but it’s a lot of meetings about a lot of different topics. And I find that one of the best parts of the job is the breadth of topics that my group tackles.
Jim Freeze It’s interesting as everybody associates Moderna today with vaccines and in particular, one of the most pervasive vaccines, the COVID 19 vaccine that you’ve developed and it has been so impactful. But the company actually has its roots as a biotech startup and developed its entire drug discovery and manufacturing process around artificial intelligence. Can you share a bit more about this approach and its impact on Moderna’s vaccine initiatives?
Andrew Geissel Yeah, sure. To me, one of the most compelling things about Moderna, and I think it’s true for many people, both inside and outside of the company, is that it really is a platform. The difference between say a vaccine for flu and a vaccine for COVID is the sequence of the mRNA. And so the equipment is the same. The input goods are the same, the digital processes are all the same. You just change a text string at the very beginning of the process. And so, our CEO liked to joke that we would have no products or 1,000 products because if it works once, it really does a lot to give you confidence that it’ll work again.
So I think with that in mind, we’ve always been a digital company. The very essence of our products is stored in a computer as a string of letters. That has been really enabling for a lot of different processes, a lot of different projects at the company. Vaccines are one very powerful application of the technology, but there’s many others: enzyme replacement, gene editing, oncology, and we’ve had clinical trials in all these different areas. So when we tackle problems at the company, I think we try to have a platform mindset. We don’t want to solve a problem just once. We want to solve a problem that solves it over and over again, basically. And so for some of the pieces that my team was involved in with the COVID vaccine development actually weren’t necessarily developed for the COVID vaccine and there’s three different pieces I guess I would highlight.
One is to make mRNA, you start with the DNA template and we design the sequence of the DNA template and thus the sequence of the mRNA and then the protein that is eventually made. To do that, we buy DNA from companies just like labs around the world. And you generally can’t buy a full link gene at a reasonable price. You usually have to buy them in smaller pieces and then combine them in the lab and make them and then do a quality control step to make sure that you combine them in the right way.
So we have had algorithms that are involved in that process in two different ways. One is an algorithm that finds the best place to break apart the gene so that we have the best chance of reassembling it into the correct sequence on the first shot. And then there’s a secondary quality control step where we look at a part of the gene that is at the very tail end. It’s a whole bunch of the letter A, so it’s called a poly-A tail. Analyzing that or making sure that that was correct is very difficult for computers to do in an easy way, but humans can do it well.
And so we partnered with our colleagues that did that step to label a lot of examples of tails that passed and failed, and trained an algorithm to be able to predict the probability of the tail being high quality or not. So the way that that was deployed is in an app that the app engineering team has built that when they get the results back from the quality control, from our vendor basically, they’re ranked by this score. And that helps them narrow in on the correct sequence to pass on to the next step.
Jim Freeze Wow.
Andrew Geissel So it is a real back and forth of scientists. We depended on their expertise to label the data. And then the algorithm wasn’t necessarily just employed in a replacement sort of way. It’s deployed as a tool to help scientists do their job better. Both of these were developed far before the pandemic, but they aided in our ability to make the first batch of material for the phase one clinical trials as quickly as possible.
Jim Freeze So I hear that Moderna is doing quite a bit to get everyone at the company trained in some capacity about artificial intelligence so that they can leverage it and understand it in day-to-day operations. Can you tell us a little bit more about this?
Andrew Geissel One of the things we’re doing at Moderna is thinking hard about how we want to use AI across the company. And I think for many people, that’s going to take a little bit of a change in the way that they think or the background that they have. And so we are about to launch an academy where we are going to essentially train every single employee with a multiple-day workshop on the basics of AI, statistical thinking, data collection.
Jim Freeze That’s fantastic. That’s great.
Andrew Geissel Yeah. It’s pretty remarkable. I think when we looked around for education-like programs, they’re basically stuff for practitioners and stuff for executives and nothing in between. And so we are building this three-tier course. The first level is essentially the entire company is like AI awareness training. And then there’s an AI in practice second tier, which I consider stakeholder education. It’s like the things that I would do when meeting with the new potential stakeholder to talk to them about, does their problem fit into, say a machine learning context, specification gathering in a software engineering sense, but on the data science side.
So, a good chunk of the company, maybe upwards of 40%, 50% of the company is going to go through that. And so the hope is that at every single level, if anyone recognizes an opportunity, they can capture it and they’ll have the vocabulary with which to talk about it. And the knowledge of what is necessary for a successful machine learning artificial intelligence project. Then we’ll have a third tier, which is basically going to be about business practices like business and AI, change management, and how to evaluate ROI around these things. And then a fourth tier, which is for the very senior people in the company to think at a very high strategic level about how they’re part of the company is going to be shaped by changes in AI over the future. It’s a tremendous investment.
Jim Freeze What’s really interesting about that is one of the first episodes we did, I think it was in season one was we talked with, I forget who it was. I have to think back to the episode, but we got into a discussion around education and specifically high school education and talking about how antiquated our education system is based on what we teach and more specifically what we don’t teach.
And the reference was to, why isn’t every high school student learning statistics? Because of how critical that is to the economy and where we’re heading from an artificial intelligence perspective. So it’s really interesting to see a company like Moderna just embracing it and making sure that all employees at all levels are educated about AI. I think that’s great.
So beyond vaccine development, do you see trends or changes in the way AI was used or is being used in the pharmaceutical industry specifically as a result of the pandemic?
Andrew Geissel That’s interesting to think about it as a result of the pandemic. I can say that at Moderna, the pandemic and our ability to make an effective vaccine has changed a lot of things about Moderna. And we are hoping to integrate AI into all sorts of business processes. We need to be as efficient as possible if we want to make the next vaccine or the next 10 vaccines, or the next 100. So, our group, aside from these applications in the manufacturing process, we’re also using AI for schedule optimization, for in vitro studies, forecasting adverse events and adverse event rates, clinical trial design, epidemiology studies, protein engineering and design, and natural text document generation. I don’t think there is a part of business that is not touched by AI if your company is digital enough.
The most successful pharmaceutical companies are realizing that and investing in it. In terms of other changes, industry-wide that have been driven by the pandemic, I think that generally speaking, mRNA as a modality, or as a type of therapy, I think is of great interest. I think for a long time people weren’t sure if mRNA therapies were going to be as effective as they have shown to be. To the extent that AI can aid in mRNA drug development, I think we’re seeing many different startups that are all tackling problems in this space and they all have opportunities to apply AI.
Jim Freeze It’s interesting to hear you talk about AI, it pretty much impacts every segment or every portion of your business. The pharma industry in general has seen just tremendous amounts of innovation over the last 18 months, but I’m confident that there’s more ahead. The next 5 to 10 years, what other areas in medicine or pharma stand to benefit from AI?
Andrew Geissel Yeah, that’s a great one. I think likely one of the first major use cases for AI is in image analysis, probably speaking, it’s one of the most developed domains. And in the medical domain, that typically is radiology images and also pathology images. There are many great startups that are working on these problems. And I think it harkens back to what we had. The way that we needed to work with our colleagues, to put the results of our model in a way that they could use it to help make better decisions. And I think that that’s probably the way that feels like digital radiology and histopathology are going to go.
They’re not going to replace people, but they’re going to give them other angles on their data or make the tedious parts of their job less tedious. But it’ll be a partnership between these algorithms that have been trained on more data than a human could ever see. And a human who has a deep knowledge of the world in medicine that is to date not been embedded in a computer. So this human experience and just massive data processing capabilities of AI, I think together can really do wonders.
Jim Freeze Yeah. Well, philosophically, I work for a company that fundamentally believes that humans need AI and AI needs humans. You’ve just actually emphasized that point with a big exclamation point. Thank you so much.I’ve learned a ton and boy, if there’s ever been an episode that highlights the positive impact that artificial intelligence is having on society and our lives, this has been it. We really appreciate the time that you took today to walk us through your experience at Moderna and how AI is really changing pharma and medicine. Thank you.
Andrew Geissel Yeah, my pleasure.
That’s a wrap for this episode—and this season—of the ConversAItion. A big thank you to all of our guests, and to all of our listeners for tuning in. We’ll be back in the spring with more timely discussions from leading AI thought leaders, business people and academics.
This episode of The ConversAItion podcast was produced by Interactions, a conversational AI company. I’m Jim Freeze, and we’ll see you next season.