Introduced as Korea’s first commercial software based on Reinforcement Learning, the pinnacle of global AI technology.
Technology verified in the financial industry…expanding and applying to manufacture, public, education, and distribution.

On the 5th, AI specialized AgileSoDA (CEO, Choi Daewoo) announced that it has launched ‘BakingSoDA’, an Agent Maker based on Reinforcement Learning technology. BakingSoDA is an enterprise AI software that approaches the “optimization” problem which has gained global attention due to Reinforcement Learning. Commercial products that introduce Reinforcement Learning is a rare case in the global market.
 ‘Optimization’ is the process of finding a closer answer to the desired goal in a given circumstances. Recently, ‘optimization’ has been spotlighted as an ideal field of AI technology since it can reduce corporate expenses and maximizes profit.
 AgileSoDA tries to solve this optimization issue with Reinforcement Learning. Reinforcement Learning is a methodology that establishes a data-based optimal strategy. AgileSoDA has demonstrated its excellent performance by incorporating Reinforcement Learning that optimizes financial credit limits, detecting fraudulent activities, and improving design modules.
However, the design and application of Reinforcement Learning to optimization has been extremely rare since it’s too difficult to complete. This is because it requires high expertise in both algorithms and relevant fields, and has to go through trial and error such as setting an appropriate reward that fits the goals.
AgileSoDA overcame the difficulties that analysts are experiencing and made Reinforcement Learning easier. Analysts can use BakingSoDA’s user-friendly interface (UI) and analysis Pipe-line flow to develop Reinforcement Learning agents in a click-by-click form.
BakingSoDA provides a convenient reward setting. This reward setting is selectively supported by an automatic function, a semi-automated function through wizard function, and an arbitrary setting function according to the user’s need. A paper on ‘A2GAN’, AgileSoDA’s self-developed algorithm for automating reward settings, was adopted by NeurIPS. This is the first case in the world to implement this reward automation function as a commercial product.
“BakingSoDA is a product of technology and knowhow that realizes ‘optimization’ and ‘automation’. It is the product that represents the core value of AgileSoDA. We are working on various projects including finance, manufacture, education, and distribution. In the future, it will be combined with various systems and fields. It will be used more widely,” said Choi Daewoo, CEO of AgileSoDA.