Startup Interview
- [Startup Interview] CTO Tae-Ho Kim of “Nota”, an On-device AI company based on deep learning model compression technology
- 창업원
- 2022-10-30 20:27:02
- 382
Today we will meet CTO Tae-Ho Kim of Nota, a company with a motto of creating a convenient world using AI.
The AI model compression technology developed by Nota supports a variety of products, and is cost and time-efficient for product and service development. Through an automatic AI compression platform called NetsPresso, Nota expands into a business utilizing AI and compression solutions. Let’s meet CTO Tae-Ho Kim of Nota!
Can you briefly introduce Nota?
Hello, Nota develops and researches deep learning model compression technologies to help other companies use their AI models more efficiently. We help minimize costs for customers with AI models on servers, and enable customers operating AI models on edges to use cheaper but higher performance models on the same edge. We also develop compression AI models such as face authentication and vehicle detection.
Deep learning compression technology for AI operated on edges like smartphones and drones will help efficient operation on other devices as well. Using Nota’s technology, compression technology that does not rely on high-performance servers will reduce costs in various aspects.
Why did Nota decide on this item as its startup product?
We initially did not start with this specific item. Our first startup item was an AI solution on smartphone soft keyboards that helps reduce typos. During the solution development process, we realized the difficulty of running AI on devices, and decided to pivot to solve the chronic issue of heavy AI models.
Although the item is different, it seems to be still based on AI. Nota is one of the early startup teams from the student startup audition program E*5. Since then, Nota has drawn a blueprint for artificial intelligence, which has led to the current AI deep learning model compression technology!
Recently, the number of AI companies is increasing. What is unique about Nota, and what advantages does the company have?
First, I believe that most AI companies develop AI solutions in specific fields. Nota has these AI companies as customers and help them streamline AI models through our AI compression technology. As the demand and solution for AI increase, the experience of Nota and our platform NetsPresso will help people wanting to apply AI in the real world.
Unlike other companies that develop AI solutions, Nota collaborates with other companies, pursues efficiency of AI models, and develops an automated AI model compression platform called NetsPresso.
Can you explain more about Nota’s automated AI model compression platform ‘NetsPresso’?
NetsPresso is an automated deep learning model compression platform. If users input their deep learning model, NetsPresso compresses the input model into a lighter model. NetsPresso receives AI models, datasets, target HW, and the customers’ requirements as the input.
NetsPresso consists largely of a compression technology pool, device pool, optimization algorithms, and compression pipelines. A compression technology pool includes various compression technologies like Network Pruning and Quantization, and other technologies that Nota is developing. NetsPresso selects various compression technologies in the pool to an input model, and the optimization algorithms decide how and which compression technology to use. When the algorithm determines the technologies, it chooses the optimal combination of compression technologies to the target HW inputted by the customer. Then, the device pool builds the model onto the connected target HW and measures the performance in that HW.
NetsPresso can solve the problems of heavy AI models and computations customers face, and reduce costs to customers. NetsPresso aims to deliver realistic values such as enabling AI on-devices.
NetsPresso is the optimized platform to apply compression technologies to AI devices. It can solve existing problems and increase efficiency by applying various compression technologies to input models through optimization algorithms. With NetsPresso, anyone can easily and cheaply compress AI models, which explains why any companies, including domestic large corporations, form partnerships with Nota.
What does Nota wish to establish through AI model compression and on-device AI solutions?
AI’s capabilities and growth are explosive, but there are many barriers until AI completely integrates with our lives, the main reason being the heavy size and computations of AI models. Through AI model compression and on-device AI solutions, we want AI to have positive effects on our lives.
AI’s growth rate is extremely fast and dynamic. In the midst, there are still challenges to be solved, but we expect NetsPresso to solve compression technology issues and reduce the time and costs involved.
What are the difficulties in compressing AI models, how does Nota overcome them?
AI models are extremely large and has a lot of computations. To compress AI models, there are various methods such as Network Pruning, Quantization, Knowledge Distillation, and Filter Decomposition. Each technique is mainly studied in academia. Thus, to utilize these techniques in the industry for compression, a deep understanding of the model and know-hows of appropriately utilizing techniques depending on the given model and dataset are required. NetsPresso is an automated compression platform that helps effectively compress models, which was previously done by humans. As I explained about NetsPresso earlier, NetsPresso can automate the compression process and make it easier and more effective.
Compressing AI model sounds like a difficult task. Nevertheless, Nota succeeded in developing an automated platform, NetsPresso, that has replaced human work, making the compression process more convenient.
Nota is known to collaborate with other companies. What kind of collaborations are there and how is Nota’s technology being used?
Nota’s business areas can be divided into three main categories – the compression business, face authentication solution, and an intelligent transportation system business that uses compressed solutions. For businesses that uses compression technologies, we collaborate with companies that already have AI models or are planning on commercialization through AI models. For example, a company that uses AI models on edge devices such as drones or AI cameras can use lighter equipment and more camera channels through AI model compression. In addition, we collaborate with companies that use cloud servers to service AI to help reduce the cloud GPU usage.
Other than compression businesses, we also provide solutions using compression technologies. Our face authentication solution helps manage entrance to construction sites, and buildings. You can check out more collaboration businesses through the link below.
It seems that Nota’s technology is competitive enough without the collaboration with other companies. We can see that compression technologies can be utilized in many businesses, and is already widely used. Nota’s technology has positive impact on our lives, including reducing traffic jams. (Please click the link above to see how the technology is used in the real world)
There must be a change in the environment since Nota first prepared to start a business. What are some differences, and what advice can you give to KAIST students who want to start a business?
Nota was established in 2015. As a graduate student, I focused on research, but I saw that the numerous research, including mine, was limited to academia. I wanted to directly solve the problems in society and decided to start a business. During the startup process, we experienced difficulties while pivoting, but overcame them with support of team members and the thought that we could still solve the problems in the society. When you are experiencing hardships during the startup process, remember why you started this journey.
It’s been a while since Nota first started the business. The success of Nota lies in the potential in AI and the team members. Unlike other AI companies, Nota went through pivoting, but successfully created an automated AI model compression platform. Paying attention to what the society and world needs is a good starting point in starting a business!
What is Nota’s ultimate goal and vision?
There is a huge difference between life 10 years ago and life today. I believe there will be a big difference in life today and 10 years from now as well. In an era of explosive growth in artificial intelligence, impacting people through technology is touching. When AI is projected into our lives and society and causes various changes, I want to solve the problem of size and computation of models that become barriers to these changes. Through this, I wish AI enriches people’s lives.
Who knew that the AI era would come so rapidly? Just two years ago, no one expected COVID-19 to invade our lives. When AI was first introduced, people thought that AI could danger our lives, but now there isn’t a device that does not use AI. In this era of AI, the company Nota not only provides AI solutions but also develops AI model compression technology that can help these solutions. With its groundbreaking AI model compression technology, Nota is attracting attention around the world and has also been introduced at CES 2020 as well as CES 2021. We look forward to the prospective work of Nota.
This was CTO Tae-Ho Kim of Nota.
We initially did not start with this specific item. Our first startup item was an AI solution on smartphone soft keyboards that helps reduce typos. During the solution development process, we realized the difficulty of running AI on devices, and decided to pivot to solve the chronic issue of heavy AI models.
Although the item is different, it seems to be still based on AI. Nota is one of the early startup teams from the student startup audition program E*5. Since then, Nota has drawn a blueprint for artificial intelligence, which has led to the current AI deep learning model compression technology!
First, I believe that most AI companies develop AI solutions in specific fields. Nota has these AI companies as customers and help them streamline AI models through our AI compression technology. As the demand and solution for AI increase, the experience of Nota and our platform NetsPresso will help people wanting to apply AI in the real world.
Unlike other companies that develop AI solutions, Nota collaborates with other companies, pursues efficiency of AI models, and develops an automated AI model compression platform called NetsPresso.
Can you explain more about Nota’s automated AI model compression platform ‘NetsPresso’?
NetsPresso is an automated deep learning model compression platform. If users input their deep learning model, NetsPresso compresses the input model into a lighter model. NetsPresso receives AI models, datasets, target HW, and the customers’ requirements as the input.
NetsPresso consists largely of a compression technology pool, device pool, optimization algorithms, and compression pipelines. A compression technology pool includes various compression technologies like Network Pruning and Quantization, and other technologies that Nota is developing. NetsPresso selects various compression technologies in the pool to an input model, and the optimization algorithms decide how and which compression technology to use. When the algorithm determines the technologies, it chooses the optimal combination of compression technologies to the target HW inputted by the customer. Then, the device pool builds the model onto the connected target HW and measures the performance in that HW.
NetsPresso can solve the problems of heavy AI models and computations customers face, and reduce costs to customers. NetsPresso aims to deliver realistic values such as enabling AI on-devices.
NetsPresso is the optimized platform to apply compression technologies to AI devices. It can solve existing problems and increase efficiency by applying various compression technologies to input models through optimization algorithms. With NetsPresso, anyone can easily and cheaply compress AI models, which explains why any companies, including domestic large corporations, form partnerships with Nota.
What does Nota wish to establish through AI model compression and on-device AI solutions?
AI’s capabilities and growth are explosive, but there are many barriers until AI completely integrates with our lives, the main reason being the heavy size and computations of AI models. Through AI model compression and on-device AI solutions, we want AI to have positive effects on our lives.
AI’s growth rate is extremely fast and dynamic. In the midst, there are still challenges to be solved, but we expect NetsPresso to solve compression technology issues and reduce the time and costs involved.
What are the difficulties in compressing AI models, how does Nota overcome them?
AI models are extremely large and has a lot of computations. To compress AI models, there are various methods such as Network Pruning, Quantization, Knowledge Distillation, and Filter Decomposition. Each technique is mainly studied in academia. Thus, to utilize these techniques in the industry for compression, a deep understanding of the model and know-hows of appropriately utilizing techniques depending on the given model and dataset are required. NetsPresso is an automated compression platform that helps effectively compress models, which was previously done by humans. As I explained about NetsPresso earlier, NetsPresso can automate the compression process and make it easier and more effective.
Compressing AI model sounds like a difficult task. Nevertheless, Nota succeeded in developing an automated platform, NetsPresso, that has replaced human work, making the compression process more convenient.
Nota is known to collaborate with other companies. What kind of collaborations are there and how is Nota’s technology being used?
Nota’s business areas can be divided into three main categories – the compression business, face authentication solution, and an intelligent transportation system business that uses compressed solutions. For businesses that uses compression technologies, we collaborate with companies that already have AI models or are planning on commercialization through AI models. For example, a company that uses AI models on edge devices such as drones or AI cameras can use lighter equipment and more camera channels through AI model compression. In addition, we collaborate with companies that use cloud servers to service AI to help reduce the cloud GPU usage.
Other than compression businesses, we also provide solutions using compression technologies. Our face authentication solution helps manage entrance to construction sites, and buildings. You can check out more collaboration businesses through the link below.
It seems that Nota’s technology is competitive enough without the collaboration with other companies. We can see that compression technologies can be utilized in many businesses, and is already widely used. Nota’s technology has positive impact on our lives, including reducing traffic jams. (Please click the link above to see how the technology is used in the real world)
There must be a change in the environment since Nota first prepared to start a business. What are some differences, and what advice can you give to KAIST students who want to start a business?
Nota was established in 2015. As a graduate student, I focused on research, but I saw that the numerous research, including mine, was limited to academia. I wanted to directly solve the problems in society and decided to start a business. During the startup process, we experienced difficulties while pivoting, but overcame them with support of team members and the thought that we could still solve the problems in the society. When you are experiencing hardships during the startup process, remember why you started this journey.
It’s been a while since Nota first started the business. The success of Nota lies in the potential in AI and the team members. Unlike other AI companies, Nota went through pivoting, but successfully created an automated AI model compression platform. Paying attention to what the society and world needs is a good starting point in starting a business!
What is Nota’s ultimate goal and vision?
There is a huge difference between life 10 years ago and life today. I believe there will be a big difference in life today and 10 years from now as well. In an era of explosive growth in artificial intelligence, impacting people through technology is touching. When AI is projected into our lives and society and causes various changes, I want to solve the problem of size and computation of models that become barriers to these changes. Through this, I wish AI enriches people’s lives.
Who knew that the AI era would come so rapidly? Just two years ago, no one expected COVID-19 to invade our lives. When AI was first introduced, people thought that AI could danger our lives, but now there isn’t a device that does not use AI. In this era of AI, the company Nota not only provides AI solutions but also develops AI model compression technology that can help these solutions. With its groundbreaking AI model compression technology, Nota is attracting attention around the world and has also been introduced at CES 2020 as well as CES 2021. We look forward to the prospective work of Nota.
This was CTO Tae-Ho Kim of Nota.
NetsPresso is an automated deep learning model compression platform. If users input their deep learning model, NetsPresso compresses the input model into a lighter model. NetsPresso receives AI models, datasets, target HW, and the customers’ requirements as the input.
NetsPresso consists largely of a compression technology pool, device pool, optimization algorithms, and compression pipelines. A compression technology pool includes various compression technologies like Network Pruning and Quantization, and other technologies that Nota is developing. NetsPresso selects various compression technologies in the pool to an input model, and the optimization algorithms decide how and which compression technology to use. When the algorithm determines the technologies, it chooses the optimal combination of compression technologies to the target HW inputted by the customer. Then, the device pool builds the model onto the connected target HW and measures the performance in that HW.
NetsPresso can solve the problems of heavy AI models and computations customers face, and reduce costs to customers. NetsPresso aims to deliver realistic values such as enabling AI on-devices.
NetsPresso is the optimized platform to apply compression technologies to AI devices. It can solve existing problems and increase efficiency by applying various compression technologies to input models through optimization algorithms. With NetsPresso, anyone can easily and cheaply compress AI models, which explains why any companies, including domestic large corporations, form partnerships with Nota.
AI’s capabilities and growth are explosive, but there are many barriers until AI completely integrates with our lives, the main reason being the heavy size and computations of AI models. Through AI model compression and on-device AI solutions, we want AI to have positive effects on our lives.
AI’s growth rate is extremely fast and dynamic. In the midst, there are still challenges to be solved, but we expect NetsPresso to solve compression technology issues and reduce the time and costs involved.
What are the difficulties in compressing AI models, how does Nota overcome them?
AI models are extremely large and has a lot of computations. To compress AI models, there are various methods such as Network Pruning, Quantization, Knowledge Distillation, and Filter Decomposition. Each technique is mainly studied in academia. Thus, to utilize these techniques in the industry for compression, a deep understanding of the model and know-hows of appropriately utilizing techniques depending on the given model and dataset are required. NetsPresso is an automated compression platform that helps effectively compress models, which was previously done by humans. As I explained about NetsPresso earlier, NetsPresso can automate the compression process and make it easier and more effective.
Compressing AI model sounds like a difficult task. Nevertheless, Nota succeeded in developing an automated platform, NetsPresso, that has replaced human work, making the compression process more convenient.
Nota is known to collaborate with other companies. What kind of collaborations are there and how is Nota’s technology being used?
Nota’s business areas can be divided into three main categories – the compression business, face authentication solution, and an intelligent transportation system business that uses compressed solutions. For businesses that uses compression technologies, we collaborate with companies that already have AI models or are planning on commercialization through AI models. For example, a company that uses AI models on edge devices such as drones or AI cameras can use lighter equipment and more camera channels through AI model compression. In addition, we collaborate with companies that use cloud servers to service AI to help reduce the cloud GPU usage.
Other than compression businesses, we also provide solutions using compression technologies. Our face authentication solution helps manage entrance to construction sites, and buildings. You can check out more collaboration businesses through the link below.
It seems that Nota’s technology is competitive enough without the collaboration with other companies. We can see that compression technologies can be utilized in many businesses, and is already widely used. Nota’s technology has positive impact on our lives, including reducing traffic jams. (Please click the link above to see how the technology is used in the real world)
There must be a change in the environment since Nota first prepared to start a business. What are some differences, and what advice can you give to KAIST students who want to start a business?
Nota was established in 2015. As a graduate student, I focused on research, but I saw that the numerous research, including mine, was limited to academia. I wanted to directly solve the problems in society and decided to start a business. During the startup process, we experienced difficulties while pivoting, but overcame them with support of team members and the thought that we could still solve the problems in the society. When you are experiencing hardships during the startup process, remember why you started this journey.
It’s been a while since Nota first started the business. The success of Nota lies in the potential in AI and the team members. Unlike other AI companies, Nota went through pivoting, but successfully created an automated AI model compression platform. Paying attention to what the society and world needs is a good starting point in starting a business!
What is Nota’s ultimate goal and vision?
There is a huge difference between life 10 years ago and life today. I believe there will be a big difference in life today and 10 years from now as well. In an era of explosive growth in artificial intelligence, impacting people through technology is touching. When AI is projected into our lives and society and causes various changes, I want to solve the problem of size and computation of models that become barriers to these changes. Through this, I wish AI enriches people’s lives.
Who knew that the AI era would come so rapidly? Just two years ago, no one expected COVID-19 to invade our lives. When AI was first introduced, people thought that AI could danger our lives, but now there isn’t a device that does not use AI. In this era of AI, the company Nota not only provides AI solutions but also develops AI model compression technology that can help these solutions. With its groundbreaking AI model compression technology, Nota is attracting attention around the world and has also been introduced at CES 2020 as well as CES 2021. We look forward to the prospective work of Nota.
This was CTO Tae-Ho Kim of Nota.
AI models are extremely large and has a lot of computations. To compress AI models, there are various methods such as Network Pruning, Quantization, Knowledge Distillation, and Filter Decomposition. Each technique is mainly studied in academia. Thus, to utilize these techniques in the industry for compression, a deep understanding of the model and know-hows of appropriately utilizing techniques depending on the given model and dataset are required. NetsPresso is an automated compression platform that helps effectively compress models, which was previously done by humans. As I explained about NetsPresso earlier, NetsPresso can automate the compression process and make it easier and more effective.
Compressing AI model sounds like a difficult task. Nevertheless, Nota succeeded in developing an automated platform, NetsPresso, that has replaced human work, making the compression process more convenient.
Nota’s business areas can be divided into three main categories – the compression business, face authentication solution, and an intelligent transportation system business that uses compressed solutions. For businesses that uses compression technologies, we collaborate with companies that already have AI models or are planning on commercialization through AI models. For example, a company that uses AI models on edge devices such as drones or AI cameras can use lighter equipment and more camera channels through AI model compression. In addition, we collaborate with companies that use cloud servers to service AI to help reduce the cloud GPU usage.
Other than compression businesses, we also provide solutions using compression technologies. Our face authentication solution helps manage entrance to construction sites, and buildings. You can check out more collaboration businesses through the link below.
It seems that Nota’s technology is competitive enough without the collaboration with other companies. We can see that compression technologies can be utilized in many businesses, and is already widely used. Nota’s technology has positive impact on our lives, including reducing traffic jams. (Please click the link above to see how the technology is used in the real world)
There must be a change in the environment since Nota first prepared to start a business. What are some differences, and what advice can you give to KAIST students who want to start a business?
Nota was established in 2015. As a graduate student, I focused on research, but I saw that the numerous research, including mine, was limited to academia. I wanted to directly solve the problems in society and decided to start a business. During the startup process, we experienced difficulties while pivoting, but overcame them with support of team members and the thought that we could still solve the problems in the society. When you are experiencing hardships during the startup process, remember why you started this journey.
It’s been a while since Nota first started the business. The success of Nota lies in the potential in AI and the team members. Unlike other AI companies, Nota went through pivoting, but successfully created an automated AI model compression platform. Paying attention to what the society and world needs is a good starting point in starting a business!
What is Nota’s ultimate goal and vision?
There is a huge difference between life 10 years ago and life today. I believe there will be a big difference in life today and 10 years from now as well. In an era of explosive growth in artificial intelligence, impacting people through technology is touching. When AI is projected into our lives and society and causes various changes, I want to solve the problem of size and computation of models that become barriers to these changes. Through this, I wish AI enriches people’s lives.
Who knew that the AI era would come so rapidly? Just two years ago, no one expected COVID-19 to invade our lives. When AI was first introduced, people thought that AI could danger our lives, but now there isn’t a device that does not use AI. In this era of AI, the company Nota not only provides AI solutions but also develops AI model compression technology that can help these solutions. With its groundbreaking AI model compression technology, Nota is attracting attention around the world and has also been introduced at CES 2020 as well as CES 2021. We look forward to the prospective work of Nota.
This was CTO Tae-Ho Kim of Nota.
Nota was established in 2015. As a graduate student, I focused on research, but I saw that the numerous research, including mine, was limited to academia. I wanted to directly solve the problems in society and decided to start a business. During the startup process, we experienced difficulties while pivoting, but overcame them with support of team members and the thought that we could still solve the problems in the society. When you are experiencing hardships during the startup process, remember why you started this journey.
It’s been a while since Nota first started the business. The success of Nota lies in the potential in AI and the team members. Unlike other AI companies, Nota went through pivoting, but successfully created an automated AI model compression platform. Paying attention to what the society and world needs is a good starting point in starting a business!
There is a huge difference between life 10 years ago and life today. I believe there will be a big difference in life today and 10 years from now as well. In an era of explosive growth in artificial intelligence, impacting people through technology is touching. When AI is projected into our lives and society and causes various changes, I want to solve the problem of size and computation of models that become barriers to these changes. Through this, I wish AI enriches people’s lives.
Who knew that the AI era would come so rapidly? Just two years ago, no one expected COVID-19 to invade our lives. When AI was first introduced, people thought that AI could danger our lives, but now there isn’t a device that does not use AI. In this era of AI, the company Nota not only provides AI solutions but also develops AI model compression technology that can help these solutions. With its groundbreaking AI model compression technology, Nota is attracting attention around the world and has also been introduced at CES 2020 as well as CES 2021. We look forward to the prospective work of Nota.
This was CTO Tae-Ho Kim of Nota.