Andrew McConnell
Project Description
Landed AI explores community-owned AI that runs on owned hardware, keeps data within the community, and minimizes reliance on cloud services of the internet. Local SLMs develops and evaluates small language models (SLMs) for Indigenous language and cultural contexts, with a particular focus on Anishinaabemowin, while building the supporting infrastructure needed for community-controlled AI systems. This works advances an Urban Sovereign Stack: a self-hosted and owned ecosystem of AI models, storage, and infrastructure. The central focus of this work is Indigenous data sovereignty at both the individual and community levels, exploring how open-source tools can support the responsible stewardship of all forms of data, including sacred, cultural, and personal Indigenous knowledge. Guided by the principle of relationship as infrastructure, community-curated bundles are shared peer-to-peer within Indigenous communities and other Landed AI nodes, with governance rooted in community protocols rather than solely platform-based access controls.
A core component of these projects is education towards independence: teaching Indigenous communities to decolonize their digital tools, networks, and AI. Research activities include :running open-source, quantized AI models (LM Studio, Ollama) locally on community-owned machines; evaluating hardware and performance trade-offs (where consumer GPUs become viable) for affordable deployments; and exploring low-power, edge AI systems for smaller installations. Research also identifies strong base models that can be adapted, trained locally, and refined into SLMs to support Anishinaabemowin learning for both community members and Indigenous educators. This project evaluates how existing LLMs perform in Anishinaabemowin/Ojibwe by re-running earlier tests and reviewing Indigenous language benchmarking methods. Supporting this work is a distributed, self-hosted infrastructure that includes NAS-based local storage and synchronization systems, as well as repurposed tablets and phones that function as community-owned smart nodes.
AI Infrastructure; AI Systems; Data; Language; Students/Learners
