Sovereign Canadian Artificial Intelligence — built for privacy, precision, and purpose.
https://canxp.ai

CanXP AI is building the foundation of Canada’s sovereign AI ecosystem — ensuring that our models, data, and infrastructure remain in Canadian hands.
We develop ethical, privacy-first, and energy-efficient language models, tools, and infrastructure to empower researchers, developers, and organizations to build the next generation of Canadian AI.

We believe AI sovereignty means more than hosting data locally — it means owning the models, the methods, and the means of innovation.

A family of instruction-tuned language models trained on Canadian context, law, and culture.
Built from open architectures like Microsoft Phi-3 and extended with Canadian-specific alignment data.

• MaplePT-Mini (v1) — open model for research and experimentation
• MaplePT vLLM Plugin — enables OpenAI-compatible API serving
• Distributed training infrastructure optimized for RTX 40/30 series GPUs, promoting accessible and sustainable AI research

Our ecosystem integrates:

• vLLM — high-performance inference engine
• Supabase — secure, self-hosted data and API layer
• Docker — portable deployment for on-prem or edge inference
• Python / FastAPI / Node.js — open service frameworks for developers
• Hugging Face — transparent, collaborative model distribution

Most AI infrastructure today is foreign-owned and cloud-controlled.
We’re demonstrating that real innovation happens through open research, local collaboration, and responsible resource use - not by funding unnecessary mega-datacenters.

CanXP AI stands for:

• Sovereignty: Canadian-controlled models and infrastructure
• Accessibility: AI that runs on consumer-grade hardware
• Sustainability: Efficiency before scale
• Transparency: Open research and reproducible methods

• Distributed & edge AI training
• Ethical model alignment and Canadian context adaptation
• Low-resource fine-tuning and quantization
• Federated model evaluation
• Privacy-preserving inference and R&D collaboration

We welcome researchers, developers, and organizations who share our vision of ethical, open, and sovereign AI.
If you’d like to collaborate or integrate our models into your work, reach out at hello@canxp.ai.

All open repositories are released under permissive licenses (MIT or Apache 2.0), unless otherwise specified.
© 2025 CanXP AI — Canadian Sovereign AI Research & Innovation