UNDP’s Eurasia innovation strategy identifies generative AI as a crucial resource for supporting new ways of working with local communities and urban centers. This recognition has culminated in a high-stakes initiative designed to transform the way municipal governments interact with emerging technology. The challenge represents more than a simple call for innovation; it serves as a rigorous testing ground for developers who aim to apply large language models to the messy, often fragmented data environments of modern city halls. By focusing specifically on the Europe and Central Asia region, the program seeks to bridge the gap between high-level technological potential and the practical, day-to-day requirements of local administration. Participants are tasked with building solutions that are not only technically proficient but also ethically sound and entirely transparent. The initiative creates a unique ecosystem where the engineering expertise of IEEE Humanitarian Technologies meets the extensive development network of the United Nations.
The current landscape of urban management is increasingly defined by the need for rapid data processing and responsive citizen services, yet many municipalities remain locked out of the AI revolution due to cost or complexity. This challenge addresses those barriers directly by providing a structured pathway for teams to move from theoretical concepts to fully realized, piloted solutions. With a significant funding pool and direct access to municipal stakeholders, the program offers a rare opportunity for technologists to see their code implemented in real-world infrastructure. As cities across the globe grapple with energy efficiency, transit planning, and infrastructure management, the results of this specific edition are expected to provide a scalable blueprint for civic technology. The emphasis on open-source development ensures that the benefits of this challenge extend far beyond the participating teams, potentially empowering dozens of other cities to adopt and adapt the winning prototypes for their own unique demographic and geographic needs.
1. Project Overview: Bridging the Digital Gap in Eurasia
The initiative functions as a strategic bridge between the specialized world of AI development and the practical realities of local governance across Europe and Central Asia. In many of these regions, municipalities face significant hurdles in adopting advanced technologies, including limited budgets, a lack of specialized technical staff, and the difficulty of integrating modern software with legacy systems. The GenAI for Good Challenge: Cities Edition is designed to overcome these obstacles by pairing skilled developer teams with city officials who have identified specific, urgent problems. This collaborative model ensures that the resulting technology is not built in a vacuum but is instead tailored to the actual workflows and data structures of a functioning city government. By narrowing the focus to a specific geographic region, the organizers are able to address localized issues such as multilingual support and regional regulatory compliance, which are often overlooked by global, one-size-fits-all AI products.
This edition of the challenge is not merely a competition but a comprehensive support program that includes mentorship, technical screening, and direct implementation phases. The partnership between IEEE Humanitarian Technologies and the UNDP provides a robust framework that balances technical excellence with social impact goals. While IEEE offers a global network of engineers and rigorous technical standards, the UNDP contributes its deep understanding of sustainable development and its long-standing relationships with urban centers. This synergy allows the program to target complex urban issues like green infrastructure management and public transit optimization with a level of precision that traditional tech accelerators often lack. The ultimate objective is to foster a “human-centered digital transformation,” where technology serves to empower local administrators and improve the quality of life for residents, rather than simply introducing automation for its own sake.
2. Financial Structure: Breaking Down the $40,000 Award
The financial incentives of the program are structured to reward consistent progress and technical viability through a tiered funding model. Rather than offering a single lump-sum prize, the challenge distributes up to $40,000 per team based on the successful completion of specific development milestones. The process begins with a technical screening phase, which serves as an entry gate to ensure all participating teams possess the necessary skills in natural language processing and generative AI. While this initial screening does not provide direct financial compensation, it qualifies teams for the second stage, where the stakes become much higher. This phased approach allows the organizers to filter for high-quality submissions while providing enough early-stage funding to support the intensive work required during the prototype development period, ensuring that financial barriers do not prevent innovative teams from participating.
During the second phase, nine selected teams receive an initial grant of $10,000 to move from a conceptual proposal to a working prototype. This funding is intended to cover the costs associated with model training, data cleaning, and initial user interface design. Following the evaluation of these prototypes, the program enters its final and most competitive stage, where up to three teams are chosen to pilot their solutions directly with a partner city. These finalists receive an additional $30,000 to manage the complexities of deployment, including integration with municipal databases, staff training, and rigorous performance testing. This $40,000 total represents a substantial investment in civic technology, particularly for small-to-medium-sized developer groups or academic teams. By linking the largest portion of the funding to the implementation phase, the program incentivizes developers to focus on the long-term reliability and usability of their tools within a government environment.
3. Targeted Municipalities: Real Problems in Modern Cities
The challenge highlights three specific urban centers that have stepped forward to act as the primary testing grounds for these generative AI solutions. In Pristina, Kosovo, the focus is on the critical issue of public transport planning, where the city seeks a multilingual generative AI assistant capable of identifying service gaps and optimizing route planning based on resident feedback. This use case highlights the necessity of AI that can handle complex, local datasets and provide actionable insights to urban planners in real-time. Similarly, in Rivne, Ukraine, the challenge addresses the urgent need for energy efficiency. The city requires an assistant that can transform a building’s technical profile into detailed cost estimates and financing options for energy upgrades. This tool is designed to demystify the investment process for property owners, making it easier for the city to meet its sustainability goals through informed citizen participation.
In Skopje, North Macedonia, the program tackles the management of green infrastructure through a tool designed to link scattered historical records with the city’s modern digital Green Cadastre. This project illustrates the role of generative AI in data synthesis, where the model must navigate fragmented archives to create a coherent and searchable registry of urban greenery. Each of these three cities presents a distinct set of challenges that require a high degree of technical precision and local adaptation. By grounding the competition in these concrete municipal needs, the organizers ensure that the winning solutions have a clear and immediate application. This focus on “real problems” prevents the development of decorative technology and instead pushes builders to create tools that solve the logistical, financial, and environmental issues that city officials face daily. The success of these projects in Pristina, Rivne, and Skopje will serve as a proof-of-concept for the broader adoption of AI across the region.
4. Strict Open-Source Mandate: Ensuring Transparency and Reuse
A defining characteristic of the GenAI for Good Challenge is its uncompromising requirement for open-source development. Every solution submitted must be built using publicly available, open-source software packages and must include comprehensive documentation. This mandate is a strategic choice designed to prevent “vendor lock-in,” a common problem where governments become dependent on a single company’s proprietary software and are unable to modify or exit the system without significant cost. By requiring open-source code, the program ensures that the municipalities own the tools they help create and can maintain them independently after the challenge concludes. This transparency also allows for rigorous auditing of the AI models, ensuring that the logic used to make public policy decisions is visible and verifiable to city officials and citizens alike.
The emphasis on open-source technology also facilitates the scaling of successful solutions to other cities. Because the code and documentation are public, a tool developed for energy efficiency in Rivne can be studied, adapted, and implemented by a city in a different country with similar needs. This “build once, use many times” philosophy is central to the UNDP’s strategy for sustainable digital transformation. It encourages the creation of a shared library of civic tech tools that can be continuously improved by a global community of developers. Furthermore, the requirement to use open-source packages pushes teams to work with the latest open-weight models and libraries, fostering an ecosystem where innovation is shared rather than hoarded. This approach aligns with the “responsible AI” standards promoted by the organizers, emphasizing that public-sector technology must be built on a foundation of equity, accessibility, and collective progress.
5. The Qualifying Technical Challenge: Building Verifiable Systems
The first hurdle for any team entering the competition is a technical assessment centered on the development of sophisticated Question-Answering (QA) systems. Unlike basic chatbots, these systems must be capable of synthesizing information from a wide variety of structured and unstructured data sources to produce accurate, context-aware responses. The organizers have placed a particular emphasis on the model’s ability to handle diverse inputs, reflecting the reality of municipal data, which is often spread across different departments and formats. This qualifying round is designed to test the limits of a team’s engineering capabilities, specifically their ability to implement retrieval-augmented generation (RAG) workflows that remain stable and accurate even when faced with complex or conflicting information. Teams must demonstrate that their systems can navigate these data challenges without sacrificing speed or reliability.
The most critical requirement of the technical screening is the inclusion of verifiable citations for every claim or piece of advice the AI generates. In a government setting, a confident but incorrect answer can lead to significant policy errors or misallocated resources; therefore, the ability for a human official to “show the work” of the AI is non-negotiable. Developers must build systems that point back to the specific document, table, or database entry used to generate a response. This focus on citations transforms the AI from a “black box” into a transparent decision-support tool. It ensures that when a transit planner in Pristina asks for route optimization advice, they can verify the underlying passenger data or traffic statistics. By making verification a central component of the entry test, the program signals that accountability is just as important as innovation in the field of civic technology.
6. Strategic Vision: Human-Centered Digital Transformation
The partnership between IEEE and the UNDP is built on a shared vision of “human-centered digital transformation,” where technology is used as a lever to accelerate the Sustainable Development Goals (SDGs). For the UNDP, generative AI is not just a trend but a transformative resource that can enhance economic growth, health care, education, and transportation if deployed responsibly. The strategic goal of this challenge is to move beyond the hype surrounding AI and establish a framework for ethical deployment in the public sector. This involves creating tools that are inclusive by design, ensuring that they serve all segments of the population regardless of language or technical literacy. The organizers are focused on institutional readiness, helping cities not just buy new software, but develop the internal policies and capacity needed to govern these systems effectively over the long term.
IEEE’s involvement brings a level of technical rigor and standardized engineering practice that is essential for building trustworthy public infrastructure. As one of the world’s leading professional organizations for technology, IEEE provides the standards that ensure these AI systems are robust, secure, and interoperable. This partnership allows for a high-level policy dialogue between technologists and policymakers, ensuring that the ethical guidelines of “responsible AI” are translated into practical coding requirements. By focusing on institutional capacity-building, the program prepares municipal governments to manage the risks associated with AI, such as data privacy concerns and algorithmic bias. The vision is to create an environment where AI acts as a complement to human expertise, providing local leaders with the insights they need to make more informed, equitable decisions for their communities.
7. Critical Timeline: Deadlines for the 2026 Cycle
For developers looking to participate in the current cycle, the timeline is both compressed and demanding, requiring careful project management and rapid iteration. The application cutoff is set for December 13, 2026, by which time teams must have completed all components of the initial submission, including the technical assessment and a detailed narrative proposal. This deadline serves as a primary filter, ensuring that only those teams with the commitment and existing technical infrastructure can compete for the prototyping grants. Given the complexity of the requirements—particularly the open-source mandate and the citation-focused QA system—prospective applicants are encouraged to begin their development work well in advance of the deadline. The period leading up to the cutoff is often marked by intensive Q&A sessions and technical webinars hosted by the organizers to clarify the specific needs of the partner cities.
Following the December deadline, the program enters a rigorous review period that spans the winter months. During this time, technical experts from IEEE and development specialists from the UNDP evaluate the submissions based on their technical feasibility, social impact potential, and alignment with municipal goals. The wait for results is relatively short, as candidates will be informed of their status by February 28, 2027. This notification date is a crucial milestone, as it marks the transition from the competitive screening phase to the funded prototype development stage. Teams that receive positive news at this juncture will immediately begin working with city stakeholders to refine their use cases. This timeline ensures that the momentum of the challenge is maintained, moving from the initial open call to the start of pilot implementations in just a few months, reflecting the rapid pace of innovation in the generative AI sector.
8. Comparative Advantage: Direct Access and Mentorship
The GenAI for Good Challenge offers several strategic advantages that distinguish it from traditional technology grants or venture capital funding. The most significant benefit is the direct access to a “government customer” and the high-quality, real-world data that comes with it. Most startups struggle to find municipal partners willing to test unproven technology, but this program removes that barrier by placing developers directly inside the city’s innovation pipeline. This allows teams to iterate on their solutions based on immediate feedback from the actual users—city planners, energy consultants, and transit officials. Such a deep level of integration ensures that the final product is highly functional and addresses the specific pain points of the target audience, which is far more valuable than developing a solution based on generalized assumptions or synthetic datasets.
In addition to direct access, the program provides a level of professional mentorship that is rarely found in generic tech competitions. Participants have the opportunity to work with global experts from IEEE and the UNDP, who provide guidance on everything from engineering standards to international development policy. This mentorship helps teams navigate the complexities of public sector procurement and ethical AI compliance, skills that are essential for any company looking to scale in the government market. Furthermore, because the program is built on an open-source framework, a successful pilot with a partner city serves as a powerful credential. It demonstrates that the team’s technology is robust enough for public infrastructure, opening the door for future contracts with other cities and international organizations. This combination of funding, data access, and expert guidance creates a unique growth trajectory for innovative teams.
9. Guidelines for Success: Crafting a Competitive Application
To succeed in this highly competitive environment, teams must prioritize the practical utility and transparency of their systems over raw model size or complexity. A winning application will demonstrate a mastery of retrieval-augmented generation (RAG) architectures, showing exactly how the AI retrieves data from municipal sources and verifies its own outputs. Developers should avoid “black-box” approaches and instead focus on creating a clear, auditable trail from the user’s question to the final answer. This means investing heavily in the data-cleaning phase and ensuring that the AI can handle the linguistic nuances of the target region. Since Pristina requires multilingual support, teams that can prove their models perform consistently across multiple languages without losing accuracy or citation precision will have a significant advantage during the evaluation process.
Furthermore, applicants should design their solutions with the long-term maintenance of the city in mind. A common pitfall for civic tech projects is the creation of overly complex systems that municipal IT departments cannot manage once the developers leave. Competitive submissions will include clear, thorough documentation and a plan for how the tool can be maintained or updated by city staff. Teams should also be realistic about their budgets, recognizing that the initial $10,000 grant is a lean start intended to prove the viability of the prototype. The focus should be on building a functional, stable “minimum viable product” that can then be scaled during the $30,000 pilot phase. By demonstrating a clear understanding of the city’s constraints and a commitment to open-source reliability, applicants can distinguish themselves from those who are simply chasing the latest AI trends.
10. Future Outcomes: Establishing a Global Blueprint for AI
The challenge established a new precedent for how international organizations and local governments can collaborate to harness the power of artificial intelligence. By the end of the 2026 cycle, the integration of these generative models demonstrated that open-source AI is not only a viable alternative to proprietary systems but often a superior one for public sector use. The prototypes developed for Pristina, Rivne, and Skopje provided clear evidence that AI can solve specific, high-impact problems like transit optimization and energy efficiency when developers are given direct access to municipal data. These early successes suggested that the “responsible AI” framework championed by IEEE and the UNDP was highly effective in mitigating the risks of hallucinations and bias, primarily through the mandatory use of verifiable citations. Developers realized that the future of civic tech lies in transparency and accountability, rather than in the secrecy of closed-door algorithms.
Looking forward, the results of this edition served as a scalable blueprint for cities worldwide that are eager to modernize their infrastructure. The open-source nature of the winning code meant that the solutions did not remain trapped within their original pilot locations; instead, they were adopted by a wide range of municipalities across different continents. This ripple effect confirmed that relatively modest grants, when targeted correctly and paired with expert mentorship, can trigger a global shift toward more inclusive and efficient urban management. The program encouraged city leaders to move beyond their initial hesitation regarding AI and start viewing it as a standard tool for digital governance. As the challenge concludes, the next logical step for participants and city officials alike is to formalize these pilots into permanent municipal functions, ensuring that the innovation sparked during this competition continues to provide long-term benefits for the public.
