Building intelligent systems that shape tomorrow.
PixxieAI is an independent research & development studio focused on creating lightweight, efficient, fully custom AI systems including our flagship model Pixie, built from the ground up.
- Pixie LLM – A custom-trained, small to mid scale model built for speed, privacy, and developer freedom
- Inference Stack – High-performance serving using Go based microservices
- Pixie Chat – A friendly, fast, web based AI assistant powered by our own tech
- Token + Context Engine – Infrastructure to handle conversation memory, safety, and reasoning modules
- Dev Tools – APIs, SDKs, and utilities for integrating Pixie into apps and platforms
AI shouldn’t be locked behind massive companies.
We’re building systems that give developers ownership, control, and transparency without relying on corporate black boxes.
Languages: Go, TypeScript, Python
Model Training: PyTorch, DeepSpeed
Inference: Go + CUDA / ROCm
Infra: Docker, Kubernetes (later), Cloud + Bare Metal hybrid
Databases: PostgreSQL, Redis
APIs: REST + gRPC
We post progress, experiments, and dev logs over on our Discord and social channels.
Stay tuned for model previews, benchmarks, and open source components.
PixxieAI will release tools, datasets, and code as we grow.
Community feedback is huge for us feel free to open issues, discuss ideas, or suggest improvements.
- Website: https://pixxieai.ca
- Discord: https://pixxieai.ca/discord
- Twitter/X: https://x.com/PixxieAI
Smarter systems. Built with purpose.
