Research

Moon Tae-yoon wins World Bank Group agriculture AI hackathon

University of Seoul student Moon Tae-yoon won the World Bank Group’s agriculture track with Dure, a lightweight AI system for collective farm sales over basic phones.

University of Seoul student wins agriculture category of World Bank Group AI hackathon
University of Seoul student wins agriculture category of World Bank Group AI hackathonFirmbee / pixabay
By Sophie Lindqvist6 min read

Why it matters

  • Moon Tae-yoon won first place in the agriculture track at the Oct. 3-4 Small AI for Development Hackathon Challenge.
  • He competed alone; teams from the University of California, Berkeley, and Carnegie Mellon University placed second and third.
  • Dure coordinates collective shipments, buyer bidding and joint transportation through basic-phone text messages.
  • The system combines an ultralightweight language model with a rule-based engine to process Tetum messages.
  • GADS 2026 ran in Seoul from Oct. 19-22, where Moon demonstrated Dure and presented its architecture.

The World Bank Group’s Small AI for Development Hackathon Challenge awarded first place in its agriculture track to University of Seoul student Moon Tae-yoon at the Oct. 3-4 event.

He competed as an individual with Dure, a system that helps smallholder farmers coordinate collective sales through text messages sent from basic mobile phones. He finished ahead of teams from the University of California, Berkeley, and Carnegie Mellon University, which placed second and third.

The challenge sought practical artificial intelligence for places with limited internet access and electricity infrastructure. That operating brief mattered because smallholder farmers often struggle to negotiate favorable prices when selling individually, while formal cooperatives carry administrative and capital barriers. Dure aims to preserve collective bargaining power without requiring farmers to build a formal cooperative.

What did Moon Tae-yoon build?

Moon is a student in the Department of International Relations at the University of Seoul. He independently conceptualized and developed Dure to coordinate collective shipments, facilitate buyer bidding and arrange joint transportation through text messages.

According to the World Bank Group, the name comes from Korea’s traditional rural community organization known as “Dure,” a system of mutual assistance among farmers. The project applies that model to smallholder commerce, strengthening bargaining power through collective transactions without formal cooperative structures.

“The knowledge of development cooperation I gained while studying international relations, along with the practical experience I acquired while participating in a KOICA [Korea International Cooperation Agency] agricultural value chain project in Timor-Leste, provided a strong foundation for developing AI-based solutions to real-world challenges,” Moon said.

“The experience was particularly meaningful because I was able to go beyond simply developing technology and propose practical solutions that could be applied in the field,” he added. His development background gave the project a defined setting: a practical market problem rather than a generic technology demonstration.

Moon’s route into the project combined international-relations study with direct work on an agricultural value chain. The World Bank Group presented Dure as an applied development tool built around a specific coordination problem. Its Tetum support and basic-phone workflow reflect that field focus, rather than an attempt to build a general-purpose AI service.

The project’s design starts with a market problem Moon had studied in the field.

How does Dure work?

The World Bank Group said Dure combines an ultralightweight language model with a rule-based engine to process messages in Tetum, one of Timor-Leste’s official languages. The project was developed with agricultural communities in developing countries in mind, including coffee farmers in Timor-Leste.

Users do not need to install a separate application or maintain a constant internet connection. The design treats basic phones as the working interface, rather than adding new hardware or continuous connectivity requirements to an already difficult market.

The World Bank Group listed three core functions for Dure:

  • Coordinate collective shipments of agricultural produce.
  • Facilitate bidding by buyers.
  • Arrange joint transportation for participating farmers.

Every function happens through text messages, keeping sales coordination inside a communication tool farmers already use. The system also extends the same workflow from individual messages to a shared group process, allowing farmers to organize a sale without assembling a conventional cooperative first.

Together, these choices target users who may not be able to add expensive devices or reliable connectivity to participate in collective selling. That matches the limited-infrastructure conditions the World Bank Group set for the challenge.

In plain terms, Dure turns ordinary texts into a shared sales process.

Why does collective selling matter?

The World Bank Group said smallholder farmers often struggle to negotiate favorable prices when selling their produce individually. Formal cooperatives can overcome that weakness, but establishing one involves significant administrative procedures, initial capital requirements and the need to reach consensus among members.

Dure aims to retain the cooperative benefits while removing the requirement to form one. Farmers can coordinate a shipment, invite buyer bids and share transportation without first completing paperwork or assembling initial capital. The system makes collective action available at the point of sale, where bargaining power matters most.

The innovation is therefore practical rather than abstract. It changes the structure around a farmer’s transaction, grouping sellers before they negotiate with buyers. Collective shipments, buyer bidding and joint transport then reinforce the same coordinated sale.

The key point is practical: farmers can sell together without building a formal cooperative first.

What can Dure’s transaction data reveal?

The World Bank Group said Dure can build a database of registered farms from transaction records. It can also generate data and recommendations to support policymaking, giving the system an output beyond completed sales.

The database draws on activity already occurring through Dure’s text-message workflow. Transaction records can populate the list of registered farms, while generated recommendations can help policymakers work from structured information about agricultural activity. That links a farmer’s everyday sale with a wider planning tool.

The two features serve different users. Farmers need the messaging workflow to complete a sale, while governments receive farm data and policy recommendations. Dure therefore links commercial participation with public information, rather than treating policy support as a separate project.

The same records can support both the sale and the policy discussion around it.

How did the project move beyond the hackathon?

The World Bank Group officially invited Moon to the Global AI & Digital Summit 2026, or GADS 2026. The summit was co-hosted by the World Bank Group, Korea’s Ministry of Economy and Finance, and the Ministry of Science and ICT. It ran in Seoul from Oct. 19-22.

At an exhibition booth, Moon operated a demonstration of Dure for representatives of governments and international organizations from around the world. He also participated as a speaker in an official session, presenting the project’s technical architecture and its potential to drive innovation in agricultural markets in developing countries.

The invitation gave a university student’s project direct exposure to institutions involved in development, digital policy and finance. The summit format also served two purposes: a working demonstration for officials and a technical explanation for people assessing how the system could support agricultural markets.

Moon’s post-event role centered on demonstration and explanation rather than a reported rollout. The summit extended Dure’s audience, but it did not by itself establish day-to-day adoption among farmers or buyers.

The recognition moved Dure into public view, while field adoption remained the next test.

What will determine Dure’s next stage?

The next stage is sustained use outside a summit booth. Farmers would need to keep using the text workflow, buyers would need to respond to collective bids, and policymakers would need usable data and recommendations for the system to deliver its full potential.

Dure’s design removes two clear barriers: installing a separate application and maintaining constant internet connectivity. It cannot remove practical limits by itself, but it gives farmers and intermediaries a shared workflow that fits the infrastructure the World Bank Group asked teams to address.

For smallholder markets, the stakes are concrete. Collective shipments and joint transport can give farmers a stronger position with buyers, while farm records can give governments information about agricultural participation. Dure now has a winning hackathon project, a ministry-backed summit audience and a working demonstration. Turning those assets into regular use is the next test.

Source: feed.koreatimes.co.kr

Share this article:

More from Sophie Lindqvist

Sophie Lindqvist

Show full bio

Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.

823 articles

Related articles

  1. Seoul AI pavilion logged $3.52M in export consultations
  2. Google DeepMind Partners with Korea to Bring Frontier AI to Science
  3. World Bank Urges 'Small AI' for Developing Economies
  4. SK Telecom and Hyosung Heavy Industries partner on AI data center power
  5. Samsung and SK Join OpenAI's Stargate in Korea AI Push

« Previous articleNext article »