Skip to main content

Cookie settings

We use cookies to ensure the basic functionalities of the website and to enhance your online experience. You can configure and accept the use of the cookies, and modify your consent options, at any time.

Essential

Preferences

Analytics and statistics

Marketing

Changes at "RuralWell AI: A Hybrid AI and Physical System for Supporting Well-Being in Rural Students"

Avatar: Saiph Savage Saiph Savage

Title (English)

  • +Bridging the AI Divide: FPGA-Based Learning Companions for Rural Students in the Global South

Body (English)

  • +
  • +
    Team name
  • +
    UNAM (Universidad Nacional Autonoma de Mexico)
  • +
    Team members (First name, LAST NAME, University)
  • +
    Haziel, Álvarez González, UNAM
  • +Luis Torres Lozano, UNAM
  • +Hans Zúñiga García, UNAM
  • +Norma Elva, Chavez, UNAM
  • +Saiph, Savage, UNAM & Northeastern University
  • +
    What area does your use case primarily fall under?
  • +
    Daily life / student life / campus
  • +
    The AI use case you are working on
  • +
    We are designing a low-cost AI assistant for students in the Global South, with an initial focus on rural regions in Mexico. The system will function as a daily learning companion, supporting students with tasks such as homework assistance, basic tutoring, information access, and skill development.
  • +
  • +To ensure accessibility in low-resource settings, we plan to build the system using FPGA-based hardware, enabling efficient, low-power, and offline-capable AI inference. This hardware layer will be complemented by integration with current generative AI APIs when connectivity is available, allowing the assistant to provide richer, context-aware support.
  • +
    Why this use case matters
  • +
    Students in rural regions of the Global South often face limited access to educational resources, teachers, and digital infrastructure. While generative AI has the potential to democratize access to knowledge, most current systems assume reliable internet, high-end devices, and continuous connectivity, which excludes many learners.
  • +
  • +By leveraging FPGAs, we can create affordable, energy-efficient, and locally deployable AI systems that reduce dependence on cloud infrastructure. This approach enables edge computing, allowing students to access core functionalities even in offline or low-bandwidth environments. At the same time, selectively connecting to generative AI APIs allows the system to scale its capabilities when connectivity is available.
  • +
  • +This hybrid approach directly addresses digital inequality, supporting more equitable access to AI-powered education.
  • +
    Your team's motivation and learning objectives
  • +
    Our team is motivated by a commitment to human-centered and equitable AI, particularly in underserved communities. We aim to design technologies that augment students’ capabilities without increasing dependency or reinforcing existing inequalities.
  • +
  • +Through this project, we seek to:
  • +
  • +Learn how to design AI systems under real-world infrastructure constraints (limited connectivity, low-cost hardware)
  • +Explore the use of FPGAs for efficient on-device AI inference and understand their trade-offs compared to cloud-based systems
  • +Develop hybrid AI architectures that combine local computation with generative AI APIs
  • +Co-design with students to ensure the system reflects local needs, cultural context, and educational realities
  • +
  • +Ultimately, our goal is to build an AI assistant that is not only technically effective, but also accessible, trustworthy, and empowering for students in rural communities.
  • +
    Your initial contribution
  • +
    Situation / Context
  • +
  • +Students in rural regions of the Global South, particularly in Mexico, face persistent barriers to quality education, including limited access to teachers, learning materials, and reliable internet connectivity. While recent advances in generative AI offer new opportunities for personalized learning and knowledge access, these technologies are largely designed for contexts with stable infrastructure, high-end devices, and continuous connectivity. As a result, many of the students who could benefit the most from AI are systematically excluded from its advantages.
  • +
  • +Critical Analysis
  • +
  • +The current ecosystem of AI-driven educational tools reflects a broader pattern of infrastructural and epistemic inequality. Most systems rely on cloud-based computation, reinforcing dependency on centralized platforms and excluding low-connectivity environments. Additionally, these tools often fail to account for local cultural contexts, language variations, and everyday realities of students in rural areas.
  • +
  • +There is also a risk that introducing AI into these settings without careful design could create new forms of dependency, surveillance, or misalignment with local educational practices. Therefore, the challenge is not only to make AI accessible, but to ensure it is appropriate, empowering, and aligned with the lived conditions of students.
  • +
  • +Perspectives and Team Deliberation
  • +
  • +Within our team, we discussed several tensions:
  • +
  • +Cloud vs. Edge AI: Some argued for leveraging powerful cloud-based generative AI systems for richer capabilities, while others emphasized the need for offline-first solutions due to connectivity constraints.
  • +Performance vs. Accessibility: We debated whether to prioritize high-performance models or systems that are affordable, energy-efficient, and deployable at scale.
  • +Automation vs. Augmentation: We considered whether the AI should act as a primary tutor or as a supportive companion that augments students’ existing learning processes.
  • +
  • +Through these discussions, we converged on a hybrid approach: combining FPGA-based edge computing for local, low-cost inference with selective integration of generative AI APIs when connectivity is available. We also aligned on designing the system as a companion, not a replacement for teachers or human support systems.
  • +
  • +Proposed Contribution and Conditions for Implementation
  • +
  • +We propose the design of a low-cost, FPGA-based AI assistant that functions as a daily learning companion for students in rural regions. The system will:
  • +
  • +Operate locally on FPGA hardware to enable low-power, affordable, and offline-capable AI support
  • +Provide core functionalities such as tutoring, question answering, and skill-building without requiring constant internet access
  • +Integrate with generative AI APIs when connectivity is available to enhance capabilities
  • +Be developed through participatory design with students and educators, ensuring cultural and contextual relevance
  • +
  • +Conditions for implementation include:
  • +
  • +Access to low-cost FPGA hardware and support for deployment in rural settings
  • +Partnerships with local schools, educators, and communities for co-design and adoption
  • +Intermittent or minimal connectivity to enable periodic access to cloud-based AI services
  • +Safeguards to ensure privacy, autonomy, and responsible use, avoiding surveillance or misuse
  • +
  • +Under these conditions, this system can contribute to reducing educational inequities by making AI accessible in contexts where it is currently out of reach, while maintaining a strong focus on human-centered and context-aware design.
  • +

Confirm

Please log in

The password is too short.

Share