NetsPresso is Nota's own proprietary AI model optimization platform. Using this service, customers can quickly and efficiently obtain AI models optimized for their hardware. In addition to the convenience, NetsPresso's models are second to none, on account of the tireless efforts of our researchers and developers. Over time, users have come to expect the very best from NetsPresso, using it to solve real world problems as and when they arise.
Nevertheless, new problems and challenges are inevitable, and it is difficult for one platform to be the sole solution. To meet ever evolving demand, Nota must remain on the cutting edge of development, expanding its range of tasks and abilities. Nota's AI Application Engineers work with this in mind, finding solutions to the problems that sit outside NetsPresso's remit.
.
.
What is my current role at Nota?
As application lead, I work through problems that can't be covered by NetsPresso alone. I work in collaboration with the sales department, and I manage the development of our AI Model Factory, with an eye on developing new NetsPresso prototypes for the future. More specifically, I work to solve problems based on user requests. Depending on the issue, we either apply new AI model tasks, or we build new applications via AI modeling.
I consider how to create the most streamlined and efficient solutions that can be applied again in the future. Our work with the AI Model Factory is carried out with the aim of quickly responding to NetsPresso's customers as and when needed. In a perfect world, the insights from our work will influence the development of NetsPresso in the future.
What are the main responsibilities of an AI Application Engineer?
As I mentioned above, an AI Application Engineer must understand the problems that are difficult to solve through NetsPresso alone. And after identifying a problem, they must work to create a new solution using AI models. As we work to solve the issues, we have a secondary responsibility to try to conceive of any potential problems that may arise again in the future, and to cut them off at the head before they develop. Additionally, AI Application Engineers must work quickly. If we can identify and solve problems efficiently, it will maximize our customer satisfaction.
What does this look like in practice?
We have two primary tasks in AI Model Factory development. The first is to create a pipeline that can produce new AI models and systems as needed, to solve customer problems using real field data. AI models must successfully work in the field, they are not just a series of numbers or abstractions. We have to consider anomalies that could arise down the line, and work to mitigate them. We do this in much the same way as NetsPresso, running live simulations to identify and solve problems.
Second, we should work to improve our efficiency in building AI models and systems. One of the main advantages of NetsPresso is its rapid speed of model construction. As AI Application Engineers, we have a responsibility to try to match this, to maximize customer satisfaction. To achieve this, we try to automate the AI model process as much as possible, reducing the involvement of engineers when we can. For example, automation can take care of the AutoML aspect of modeling (e.g. Hyper-Parameter Optimization – HPO), as well as the creation of new system templates.
What are some challenges that AI Application Engineers often face?
AI is a rapidly developing field, so we encounter many problems. It is our job to remain open-minded, and to study the latest developments as much as possible. Pictured below are some books that I have found helpful recently.
Given the complexity and variety of the field, there is no one problem that will reliably reoccur. Instead, it is better to familiarize yourself with the latest AI concepts, that way you can respond to a wide array of problems as they come up. Additionally, as user hardware varies from person to person, it is necessary to develop sound theoretical knowledge of as much hardware as possible. In reality, many people lack this depth of knowledge, and can only attest to expertise in their small area of study, leading to a somewhat conservative attitude in their work. If you can go beyond this, learning and developing as you work, you can become the kind of talent that is much sought after today in the deep learning industry.
Nota Inc. NetsPresso Team Application Part | Hancheol Park
“I am a research engineer who strives to provide accessible AI solutions for real-life problems.”