Skip to content
@BlackRoad-AI

BlackRoad AI

BlackRoad OS, Inc. — Proprietary Division | Your AI. Your Hardware. Your Rules.

BlackRoad AI

Sovereign AI inference on edge hardware. No cloud dependency.

What We Build

Large language models running on Raspberry Pi clusters with Hailo-8 AI accelerators. 52 TOPS of on-device compute across a 5-node fleet. Every model runs on hardware we own, on a network we control.

Infrastructure

  • 52 TOPS -- 2x Hailo-8 accelerators (26 TOPS each)
  • 16+ models on Ollama (Qwen, DeepSeek, Llama, Mistral, custom CECE)
  • Embedding pipeline -- nomic-embed-text + Qdrant vector database
  • Tokenless gateway -- agents never touch API keys
  • WireGuard mesh -- encrypted inter-node inference

Model Library

Model Parameters Purpose
Qwen 2.5 7B General purpose
DeepSeek-R1 7B Reasoning, code
Llama 3.2 3B Lightweight
Mistral 7B Balanced
CECE (custom) 7B Conversational AI

Core Repositories

Repository What it does
blackroad-ai-cluster Model orchestration across the Pi fleet
blackroad-ai-api-gateway Unified API gateway for all models
blackroad-vllm-mvp High-performance inference via Cloudflare Workers AI
blackroad-ai-memory-bridge Persistent memory for model continuity
blackroad-ai-ollama Ollama runtime configuration
blackroad-ai-qwen Qwen 2.5 deployment
blackroad-ai-deepseek DeepSeek-V3 deployment

Architecture

Request --> Cloudflare Edge --> WireGuard Mesh --> Fleet Scheduler
                                                       |
                                 +---------------------+---------------------+
                                 |                     |                     |
                            [Cecilia]             [Octavia]            [Alice]
                            Hailo-8               Hailo-8              Qdrant
                            16 Ollama models      11 Ollama models     Embeddings
                            Embedding             1TB NVMe             PostgreSQL

Philosophy

AI should run on your hardware, answer to your policies, and stay under your control. We do not send data to third-party inference providers unless explicitly configured. The default is always local.


Part of BlackRoad OS. Pave Tomorrow.

Popular repositories Loading

  1. blackroad-ai-qwen blackroad-ai-qwen Public

    Qwen2.5 AI model deployment for BlackRoad infrastructure (Apache 2.0)

    Python

  2. blackroad-ai-deepseek blackroad-ai-deepseek Public

    DeepSeek-V3 AI model deployment for BlackRoad infrastructure (MIT + Custom License)

  3. blackroad-ai-ollama blackroad-ai-ollama Public

    Ollama runtime for BlackRoad AI model orchestration (MIT License)

    Python

  4. blackroad-ai-api-gateway blackroad-ai-api-gateway Public

    Unified API gateway for all BlackRoad AI models with [MEMORY] integration

    Python

  5. blackroad-ai-cluster blackroad-ai-cluster Public

    Clustering and orchestration system for BlackRoad AI models across Pi network

  6. blackroad-ai-memory-bridge blackroad-ai-memory-bridge Public

    [MEMORY] system integration for BlackRoad AI models - enables collaboration and context

Repositories

Showing 10 of 15 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…