Dodumpan is a custom AI development company with the mission of transforming companies into standout enterprises by assisting their digital transformation through AI. Unlike generating plausible statements with generative AI, this company aims to analyze data from each company to provide correct answers and enhance performance. Sang-kyu Park, CEO of Dodumpan, who holds a Ph.D. in engineering with over 10 years of deep learning research, personally designs custom AI tailored to each company's characteristics.
Deriving a Single Correct Answer with Custom AI Design…Sealing Revenue Leaks

Dodumpan shared a case where efficiency was improved by deriving a single correct answer through custom AI design.

CEO Sang-kyu Park stated, “We have experience in improving cost efficiency by designing custom AI for the largest chicken franchise in Korea. Surprisingly, the franchise headquarters could not identify detailed sales by menu, time, and channel (delivery app/hall/packaging) except for the total sales per branch,” adding, “The key data was left in the store's POS, and we focused on the point that the rights to that data belong to the franchisee, not the POS company. With the franchisee's consent, we designed and distributed a program that automatically aggregates POS logs from all branches with the headquarters, reconstructing sales by menu, time, region, and order channel. Previously, branch sales had to be identified through Excel and manual calls, but thanks to this program, it could be fully automated.”
He continued, “After collecting the data, we considered situational variables. We matched regional demand characteristics and sales by pulling in commercial area data (schools, PC rooms, hospitals, offices, etc.) based on portal maps to the store radius and optimized the billing structure of the delivery app's 'flag' (exposure point),” emphasizing, “Each franchisee decides where to place the store location on the delivery app map, and each flag (exposure point) costs nearly KRW 100,000. Considering cost and margin, efficient decision-making on where and how many flags to place is crucial for maximizing profit. Dodumpan focused on calculating the impact of adding and removing flags on profits to minimize advertising costs while maintaining or expanding orders.”
As a result of custom AI design based on branch log data and commercial area data, performance followed.

CEO Sang-kyu Park said, “We verified the solution before and after its introduction for one year with 11 franchisees of a chicken franchise in Seoul. Before applying the program, KRW 500 million in advertising costs was needed to increase net profit by KRW 50 million. After applying the program, the same result was achieved with KRW 200 million in advertising costs, achieving more than half the cost savings. The key is not to maximize exposure points but to retain only those that substantially contribute to profit. This efficiency was achieved with a data-based custom AI solution.”

Some may argue that the achievements of Dodumpan are possible with generative AI. Dodumpan emphasizes that only discriminative AI can derive the optimal correct answer.

CEO Sang-kyu Park stated, “Generative AI like ChatGPT or Gemini is strong in 'explanation and summary,' but has limitations in optimizing profits with its own data. In other words, it mixes 'second and third best alternatives' to produce plausible results, causing confusion,” adding, “The discriminative AI designed by Dodumpan learns only from internal company data, excluding subjectivity and evaluating with quantitative performance. As a result, it leads to reduced costs and increased performance, as seen in the chicken franchise company.”

He continued, “This method can be expanded to areas such as manufacturing, logistics, shipbuilding, mobility, healthcare, education, and tourism. For example, it can optimize the operational efficiency of tourism companies through automation of product exposure, pricing, and packaging according to weather, events, and inflow channels. It can also assist in making the most efficient decisions in emergencies, such as recognizing external objects and guiding emergency evacuation routes. In fact, we are conducting a PoC utilizing an AI optimization model to guide the fastest evacuation route when a warship is attacked with a defense company.”

Dodumpan “Providing Comprehensive Consulting to Companies Wishing to Adopt AI…Growing in Close Partnership with Companies”

Dodumpan, founded in June this year, is in urgent need of marketing and PR personnel and reference accumulation.

CEO Sang-kyu Park stated, “Dodumpan has technological capabilities and project achievements, but lacks public awareness and sales momentum. As a company founded this year, more reference accumulation is needed,” adding, “We are securing references by hiring personnel in marketing and PR fields and actively reaching out to companies to explain and prove our technology. Many companies are inquiring through word of mouth and recommendations from existing clients, so we are having busy days.”
Dodumpan announced that it would differentiate itself by providing all necessary consulting for AI adoption, in addition to supplying solutions to client companies.

CEO Sang-kyu Park stated, “Companies embarking on digital transformation find technology unfamiliar and difficult to utilize. There should be no case of just throwing a solution at such companies and ending it. Dodumpan designs internal personnel training, internal data governance establishment, and AI business roadmaps together with the company,” adding, “Thanks to this, partner recommendations continue, leading to MOUs with companies in marketing, bakery, media, and board games. We are also discussing the establishment of internal deep learning systems with large corporations. We will continue to maintain this philosophy and focus our capabilities on making companies stand out with Dodumpan's unique custom AI services. We hope for much interest in Dodumpan's journey.”

IT Donga Reporter Kim Dong-jin (kdj@itdonga.com)

Popular News