/
Designing the future, one protein at a time: A chat with Franzi
/
Designing the future, one protein at a time: A chat with Franzi

Designing the future, one protein at a time: A chat with Franzi

July 3, 2024

·

Jelle Prins

As one of the co-founders of Cradle, I've had the privilege of watching our company grow from a mere idea a bit more than two years ago, to a vibrant team of 40 people working at the cutting edge of protein design. Our journey has been marked by challenges, breakthroughs, and the collective efforts of many talented individuals. Recently, I sat down with Franzi, our very first hire, to discuss her experience at Cradle and to reflect on the exciting work she has done in the last two years, spanning from working on difficult scientific challenges to putting together IKEA furniture. I asked her how she would one day tell the story of how she got to work at Cradle, and what she did through the early years of the company.


Machine learning + nature = endless possibilities

Franzi’s career-path started many years ago in Germany, when she began studying computer science in Munich, near her typical Bavarian hometown. Computer science seemed like an unlikely choice for a young girl. But the beauty of computer science to Franzi is that, despite what cultural stereotypes would have us believe, the things that make computer science attractive are not gender-specific. 

“I know to many it looks kind of boring on the surface,” Franzi told me, laughing. “But, once you get into it, it’s really fun, there are endless possibilities and new things to learn. And that’s attractive to anyone, male or female.”

This sense of curiosity and excitement about the possibilities of technology is something we value at Cradle. As we tackle the complex challenges, we need team members who are eager to explore and learn.

Franzi's academic background, which included a master's thesis in computational neurosciences, provided a starting point for her work with us. "With human-built systems, we use machine learning to make our own inventions better, but with natural systems, we're learning to understand the amazing intelligence of nature, built over millions of years," she explained.

“We're developing tools to understand and harness the incredible complexity of biological systems. It's a field that requires a unique blend of computer science, biology, and a deep appreciation for the intricacies of nature.”


The protein generation factory

As our conversation turned to the technical aspects of our work, I asked Franzi how she would explain, in simplified words, the setup she has helped create in the last few years.

"It's really fascinating," she said with a smile. “I usually explain it to my friends as follows: imagine a virtual protein factory staffed by machine learning robots that work together as a team. The first robot, we call it the Generator, suggests new protein sequences. Then we have the Predictor robot that specializes in judging the properties of the protein, like if it's stable at different temperatures or how active it is under certain conditions."

"At the end of the line, there's a Selector robot that looks at magnitude and variance in the property predictions per generated sequence and builds the most diverse set of sequences for the lab. After that it's an iterative process - the user tests the suggested sequences in the lab, and that data is fed back to help the AI robots build better sequences in the next round."

Franzi continues: “I always emphasize that developing these AI models is more of an experimental science than a typical engineering process. The robots get retrained for every round and every project so that users get their hand tailored ML framework.” 

"I also often explain why we have our own wet-lab, even though we're not ourselves trying to bring a protein-based product to market," [Cradle charges a monthly usage fee for its software, all IP generated with the software will always be owned by the user] "We have a wet-lab so our ML team can rapidly validate model performance on real proteins and make sure we are improving the system. It's pretty cool how it all comes together."


Early successes and learning experiences

In the early days of a startup, every project is a learning experience. Franzi shared an anecdote from one of our early projects that illustrates this point well.

"We got a request from a new partner while we were on an offsite with the team," she recalled. "They had a lab deadline in a few days and offered to run some of our designed sequences. Daniel [Cradle's CTO] and I immediately got to work, and started the generation pipeline. We delivered in time but of course we had issues to fix during the offsite. Turns out, the internet bandwidth of a little hut in the black forest is not sufficient to redeploy docker images. Nevertheless, the customer was super happy about our speed - the methods they were used to would normally take weeks."

But the real surprise came when the partner's biologists reviewed the sequences. "There were some suggested mutations that really surprised them," she explained. "They were unintuitive, the kind of thing no human would ever suggest. They weren't sure if they should even spend lab resources to test them, but we were confident in our models. We convinced them to give these sequences a shot, and it turned out that these unintuitive suggestions were some of the best performing ones."

This is very significant. "When pitching Cradle, most people immediately grasp the impact of making projects go faster and cost less, but introducing diversity in the designs is just as important. So many projects hit a dead end because humans run out of ideas to reach the target performance. It's already clear that AI is going to have a very significant impact here as well."


Working together to make the best product

Given Franzi’s background, it’s no surprise that she ended up at Cradle.

“I knew from neuroscience that using machine learning to optimize proteins was going to be the perfect application,” said Franzi, reflecting on how it all began. She seemed to have immediately grasped the potential for positive impact on the world. “Machine learning or language modeling techniques can understand connections that humans struggle with. Even though nature has been optimizing proteins for millions of years, there are still endless unexplored possibilities to try and proteins to invent. This is where machine learning can speed up the search and have an impact on the physical world. Things like finding new medications, having carbon negative production pipelines or making animal free food, can all be done by finding the right proteins. ”

In the beginning, Franzi, who now works with many other people on the ML team at Cradle, was not only developing the first version of the company’s sequence generation pipeline but also acting as the company’s very first office manager. As we reminisce on the early days - which feel like long ago, but are merely 2 years in the past -, it’s obvious that she is proud of and enjoyed all the hustle that was required, not just to build the first product, but also to organize company events, set up the first office in Zurich, and even create the occasional LinkedIn post. 

Of course now things are much different: the ML team has grown, and people can mostly focus on research or refining the pipeline and making it more robust, and Franzi is no longer the company’s event coordinator. “Even so, the company culture remains similar to the early days, and the sense of community fostered by having fun together carries over into working together every day.”


A role model for women

I’m not surprised that the company’s first engineer was a woman. This is also something that is incredibly important to Franzi. As our coffee cups were running dry and we were closing up our chat for the day, Franzi adopted a serious and passionate tone.

“I admire every woman in the space because I myself know how hard it can be to power through,” Franzi said with an earnest look in her eye. “It’s really important that companies think more about it because it’s not just due to girls not signing up for computer science. There still is implicit bias and a lot of work to be done to make diverse groups feel like they belong. It is slowly moving in the right direction, and my contribution to the solution is to just try to be a good role model for other women and girls.”

“I admire every woman in the space because I myself know how hard it can be to power through”

My conversation with Franzi left a smile on my face. Her passion for her work and for supporting women in computer science was infectious. If her goal is to be a role model, she’s knocking it out of the park, and I can’t wait to see what she helps Cradle achieve in the coming months and years.

Designing the future, one protein at a time: A chat with Franzi

July 3, 2024

·

Jelle Prins

As one of the co-founders of Cradle, I've had the privilege of watching our company grow from a mere idea a bit more than two years ago, to a vibrant team of 40 people working at the cutting edge of protein design. Our journey has been marked by challenges, breakthroughs, and the collective efforts of many talented individuals. Recently, I sat down with Franzi, our very first hire, to discuss her experience at Cradle and to reflect on the exciting work she has done in the last two years, spanning from working on difficult scientific challenges to putting together IKEA furniture. I asked her how she would one day tell the story of how she got to work at Cradle, and what she did through the early years of the company.


Machine learning + nature = endless possibilities

Franzi’s career-path started many years ago in Germany, when she began studying computer science in Munich, near her typical Bavarian hometown. Computer science seemed like an unlikely choice for a young girl. But the beauty of computer science to Franzi is that, despite what cultural stereotypes would have us believe, the things that make computer science attractive are not gender-specific. 

“I know to many it looks kind of boring on the surface,” Franzi told me, laughing. “But, once you get into it, it’s really fun, there are endless possibilities and new things to learn. And that’s attractive to anyone, male or female.”

This sense of curiosity and excitement about the possibilities of technology is something we value at Cradle. As we tackle the complex challenges, we need team members who are eager to explore and learn.

Franzi's academic background, which included a master's thesis in computational neurosciences, provided a starting point for her work with us. "With human-built systems, we use machine learning to make our own inventions better, but with natural systems, we're learning to understand the amazing intelligence of nature, built over millions of years," she explained.

“We're developing tools to understand and harness the incredible complexity of biological systems. It's a field that requires a unique blend of computer science, biology, and a deep appreciation for the intricacies of nature.”


The protein generation factory

As our conversation turned to the technical aspects of our work, I asked Franzi how she would explain, in simplified words, the setup she has helped create in the last few years.

"It's really fascinating," she said with a smile. “I usually explain it to my friends as follows: imagine a virtual protein factory staffed by machine learning robots that work together as a team. The first robot, we call it the Generator, suggests new protein sequences. Then we have the Predictor robot that specializes in judging the properties of the protein, like if it's stable at different temperatures or how active it is under certain conditions."

"At the end of the line, there's a Selector robot that looks at magnitude and variance in the property predictions per generated sequence and builds the most diverse set of sequences for the lab. After that it's an iterative process - the user tests the suggested sequences in the lab, and that data is fed back to help the AI robots build better sequences in the next round."

Franzi continues: “I always emphasize that developing these AI models is more of an experimental science than a typical engineering process. The robots get retrained for every round and every project so that users get their hand tailored ML framework.” 

"I also often explain why we have our own wet-lab, even though we're not ourselves trying to bring a protein-based product to market," [Cradle charges a monthly usage fee for its software, all IP generated with the software will always be owned by the user] "We have a wet-lab so our ML team can rapidly validate model performance on real proteins and make sure we are improving the system. It's pretty cool how it all comes together."


Early successes and learning experiences

In the early days of a startup, every project is a learning experience. Franzi shared an anecdote from one of our early projects that illustrates this point well.

"We got a request from a new partner while we were on an offsite with the team," she recalled. "They had a lab deadline in a few days and offered to run some of our designed sequences. Daniel [Cradle's CTO] and I immediately got to work, and started the generation pipeline. We delivered in time but of course we had issues to fix during the offsite. Turns out, the internet bandwidth of a little hut in the black forest is not sufficient to redeploy docker images. Nevertheless, the customer was super happy about our speed - the methods they were used to would normally take weeks."

But the real surprise came when the partner's biologists reviewed the sequences. "There were some suggested mutations that really surprised them," she explained. "They were unintuitive, the kind of thing no human would ever suggest. They weren't sure if they should even spend lab resources to test them, but we were confident in our models. We convinced them to give these sequences a shot, and it turned out that these unintuitive suggestions were some of the best performing ones."

This is very significant. "When pitching Cradle, most people immediately grasp the impact of making projects go faster and cost less, but introducing diversity in the designs is just as important. So many projects hit a dead end because humans run out of ideas to reach the target performance. It's already clear that AI is going to have a very significant impact here as well."


Working together to make the best product

Given Franzi’s background, it’s no surprise that she ended up at Cradle.

“I knew from neuroscience that using machine learning to optimize proteins was going to be the perfect application,” said Franzi, reflecting on how it all began. She seemed to have immediately grasped the potential for positive impact on the world. “Machine learning or language modeling techniques can understand connections that humans struggle with. Even though nature has been optimizing proteins for millions of years, there are still endless unexplored possibilities to try and proteins to invent. This is where machine learning can speed up the search and have an impact on the physical world. Things like finding new medications, having carbon negative production pipelines or making animal free food, can all be done by finding the right proteins. ”

In the beginning, Franzi, who now works with many other people on the ML team at Cradle, was not only developing the first version of the company’s sequence generation pipeline but also acting as the company’s very first office manager. As we reminisce on the early days - which feel like long ago, but are merely 2 years in the past -, it’s obvious that she is proud of and enjoyed all the hustle that was required, not just to build the first product, but also to organize company events, set up the first office in Zurich, and even create the occasional LinkedIn post. 

Of course now things are much different: the ML team has grown, and people can mostly focus on research or refining the pipeline and making it more robust, and Franzi is no longer the company’s event coordinator. “Even so, the company culture remains similar to the early days, and the sense of community fostered by having fun together carries over into working together every day.”


A role model for women

I’m not surprised that the company’s first engineer was a woman. This is also something that is incredibly important to Franzi. As our coffee cups were running dry and we were closing up our chat for the day, Franzi adopted a serious and passionate tone.

“I admire every woman in the space because I myself know how hard it can be to power through,” Franzi said with an earnest look in her eye. “It’s really important that companies think more about it because it’s not just due to girls not signing up for computer science. There still is implicit bias and a lot of work to be done to make diverse groups feel like they belong. It is slowly moving in the right direction, and my contribution to the solution is to just try to be a good role model for other women and girls.”

“I admire every woman in the space because I myself know how hard it can be to power through”

My conversation with Franzi left a smile on my face. Her passion for her work and for supporting women in computer science was infectious. If her goal is to be a role model, she’s knocking it out of the park, and I can’t wait to see what she helps Cradle achieve in the coming months and years.

Designing the future, one protein at a time: A chat with Franzi

July 3, 2024

·

Jelle Prins

As one of the co-founders of Cradle, I've had the privilege of watching our company grow from a mere idea a bit more than two years ago, to a vibrant team of 40 people working at the cutting edge of protein design. Our journey has been marked by challenges, breakthroughs, and the collective efforts of many talented individuals. Recently, I sat down with Franzi, our very first hire, to discuss her experience at Cradle and to reflect on the exciting work she has done in the last two years, spanning from working on difficult scientific challenges to putting together IKEA furniture. I asked her how she would one day tell the story of how she got to work at Cradle, and what she did through the early years of the company.


Machine learning + nature = endless possibilities

Franzi’s career-path started many years ago in Germany, when she began studying computer science in Munich, near her typical Bavarian hometown. Computer science seemed like an unlikely choice for a young girl. But the beauty of computer science to Franzi is that, despite what cultural stereotypes would have us believe, the things that make computer science attractive are not gender-specific. 

“I know to many it looks kind of boring on the surface,” Franzi told me, laughing. “But, once you get into it, it’s really fun, there are endless possibilities and new things to learn. And that’s attractive to anyone, male or female.”

This sense of curiosity and excitement about the possibilities of technology is something we value at Cradle. As we tackle the complex challenges, we need team members who are eager to explore and learn.

Franzi's academic background, which included a master's thesis in computational neurosciences, provided a starting point for her work with us. "With human-built systems, we use machine learning to make our own inventions better, but with natural systems, we're learning to understand the amazing intelligence of nature, built over millions of years," she explained.

“We're developing tools to understand and harness the incredible complexity of biological systems. It's a field that requires a unique blend of computer science, biology, and a deep appreciation for the intricacies of nature.”


The protein generation factory

As our conversation turned to the technical aspects of our work, I asked Franzi how she would explain, in simplified words, the setup she has helped create in the last few years.

"It's really fascinating," she said with a smile. “I usually explain it to my friends as follows: imagine a virtual protein factory staffed by machine learning robots that work together as a team. The first robot, we call it the Generator, suggests new protein sequences. Then we have the Predictor robot that specializes in judging the properties of the protein, like if it's stable at different temperatures or how active it is under certain conditions."

"At the end of the line, there's a Selector robot that looks at magnitude and variance in the property predictions per generated sequence and builds the most diverse set of sequences for the lab. After that it's an iterative process - the user tests the suggested sequences in the lab, and that data is fed back to help the AI robots build better sequences in the next round."

Franzi continues: “I always emphasize that developing these AI models is more of an experimental science than a typical engineering process. The robots get retrained for every round and every project so that users get their hand tailored ML framework.” 

"I also often explain why we have our own wet-lab, even though we're not ourselves trying to bring a protein-based product to market," [Cradle charges a monthly usage fee for its software, all IP generated with the software will always be owned by the user] "We have a wet-lab so our ML team can rapidly validate model performance on real proteins and make sure we are improving the system. It's pretty cool how it all comes together."


Early successes and learning experiences

In the early days of a startup, every project is a learning experience. Franzi shared an anecdote from one of our early projects that illustrates this point well.

"We got a request from a new partner while we were on an offsite with the team," she recalled. "They had a lab deadline in a few days and offered to run some of our designed sequences. Daniel [Cradle's CTO] and I immediately got to work, and started the generation pipeline. We delivered in time but of course we had issues to fix during the offsite. Turns out, the internet bandwidth of a little hut in the black forest is not sufficient to redeploy docker images. Nevertheless, the customer was super happy about our speed - the methods they were used to would normally take weeks."

But the real surprise came when the partner's biologists reviewed the sequences. "There were some suggested mutations that really surprised them," she explained. "They were unintuitive, the kind of thing no human would ever suggest. They weren't sure if they should even spend lab resources to test them, but we were confident in our models. We convinced them to give these sequences a shot, and it turned out that these unintuitive suggestions were some of the best performing ones."

This is very significant. "When pitching Cradle, most people immediately grasp the impact of making projects go faster and cost less, but introducing diversity in the designs is just as important. So many projects hit a dead end because humans run out of ideas to reach the target performance. It's already clear that AI is going to have a very significant impact here as well."


Working together to make the best product

Given Franzi’s background, it’s no surprise that she ended up at Cradle.

“I knew from neuroscience that using machine learning to optimize proteins was going to be the perfect application,” said Franzi, reflecting on how it all began. She seemed to have immediately grasped the potential for positive impact on the world. “Machine learning or language modeling techniques can understand connections that humans struggle with. Even though nature has been optimizing proteins for millions of years, there are still endless unexplored possibilities to try and proteins to invent. This is where machine learning can speed up the search and have an impact on the physical world. Things like finding new medications, having carbon negative production pipelines or making animal free food, can all be done by finding the right proteins. ”

In the beginning, Franzi, who now works with many other people on the ML team at Cradle, was not only developing the first version of the company’s sequence generation pipeline but also acting as the company’s very first office manager. As we reminisce on the early days - which feel like long ago, but are merely 2 years in the past -, it’s obvious that she is proud of and enjoyed all the hustle that was required, not just to build the first product, but also to organize company events, set up the first office in Zurich, and even create the occasional LinkedIn post. 

Of course now things are much different: the ML team has grown, and people can mostly focus on research or refining the pipeline and making it more robust, and Franzi is no longer the company’s event coordinator. “Even so, the company culture remains similar to the early days, and the sense of community fostered by having fun together carries over into working together every day.”


A role model for women

I’m not surprised that the company’s first engineer was a woman. This is also something that is incredibly important to Franzi. As our coffee cups were running dry and we were closing up our chat for the day, Franzi adopted a serious and passionate tone.

“I admire every woman in the space because I myself know how hard it can be to power through,” Franzi said with an earnest look in her eye. “It’s really important that companies think more about it because it’s not just due to girls not signing up for computer science. There still is implicit bias and a lot of work to be done to make diverse groups feel like they belong. It is slowly moving in the right direction, and my contribution to the solution is to just try to be a good role model for other women and girls.”

“I admire every woman in the space because I myself know how hard it can be to power through”

My conversation with Franzi left a smile on my face. Her passion for her work and for supporting women in computer science was infectious. If her goal is to be a role model, she’s knocking it out of the park, and I can’t wait to see what she helps Cradle achieve in the coming months and years.

Stay in the loop

Subscribe to our newsletter to get new posts and other Cradle updates straight to your inbox. No spam :)

Built with ❤️ in Amsterdam & Zurich

Built with ❤️ in Amsterdam & Zurich