Hugging Face's logo
Activity Feed

AI & ML interests

AI Powered by You. Decentralized. Scalable. Private. Incentivized.

Recent Activity
toderian  published a model 5 days ago
ratio1/edgeguard-cypher-qwen3-4b-v0.8-feedback-validated-gguf
toderian  updated a model 5 days ago
ratio1/edgeguard-cypher-qwen3-4b-v0.8-feedback-validated-gguf
toderian  updated a Space 6 days ago
ratio1/README
View all activity

Organization Card
Community About org cards

Ratio1

Decentralized AI infrastructure for builders and partner teams.

Ratio1 helps teams build, deploy, and operate intelligent products without giving up control of their models, data paths, runtime, or economics.

Website • GitHub • Documentation

Public models • Public datasets • What we do • How we do it • Build with Ratio1 • Why teams use Ratio1

Public Models

Model Task Artifact Status Notes
ratio1/edgeguard-cypher-qwen3-4b-v0.4-gguf EdgeGuard text-to-Cypher for cybersecurity graph investigation GGUF Q4_K_M Public release Use behind an EdgeGuard read-only/schema validator. This model is intended to assist guarded graph-query workflows, not to execute arbitrary database operations.
ratio1/redmesh-cybersecqwen-4b-lora-v0.1 Defensive CVE/CWE and vulnerability triage PEFT LoRA adapter Experimental public release Adapter-only release for research and reproducibility. Not production-promoted.

Public Datasets

Dataset Task Format License Notes
ratio1/Drone_Detection_Survaillance Drone object detection and surveillance imagery Image and XML annotation files CC-BY-NC-4.0 Public computer-vision dataset with a declared train split. Dataset Viewer preview and parquet statistics may not be available for this file layout.

Positioning

Ratio1 is built for teams that want AI infrastructure they can operate, extend, and trust. We focus on practical delivery: developer tooling, distributed execution, controllable data paths, and network economics that fit production workloads instead of locking them into a single hosted stack.

What We Do

  • Help builders ship low-code and code-first AI products faster.
  • Run intelligent workloads across decentralized and edge-connected compute.
  • Support partner deployments that need clearer operational ownership and private data movement.
  • Publish reusable models, datasets, demos, and docs for the Ratio1 ecosystem.

How We Do It

  • Edge-native execution surfaces for distributed AI workloads.
  • SDKs and runtime libraries that expose real integration primitives.
  • Operator-controlled deployment patterns for privacy-sensitive environments.
  • Incentivized participation and observable infrastructure rather than black-box hosting.

Build With Ratio1

The core builder surface is open and actively evolving.

Repository Purpose
ratio1_sdk Python SDK for integrating with the Ratio1 ecosystem.
edge_node Edge execution layer for decentralized AI nodes and workloads.
naeural_core Core runtime components used across the network stack.
deeploy-dapp App-facing deployment surface for interacting with the ecosystem.
ratio1-docs Documentation source for builder onboarding, architecture, and partner material.

Why Teams Use Ratio1

  • Sovereignty: keep control over where AI runs and how data moves.
  • Elastic capacity: expand through decentralized infrastructure when centralized-only economics are a bad fit.
  • Partner readiness: support enterprise and ecosystem deployments with clearer ownership boundaries.
  • Operational visibility: work with observable primitives instead of opaque hosted abstractions.

What You Will Find Here

This Hugging Face organization is the public front door for Ratio1 assets:

  • model cards for reusable capabilities
  • dataset cards for publishable data surfaces
  • demos and static Spaces for product and partner narratives
  • documentation links back to the ratio1-docs source

Use Notes

  • Read each model or dataset card before use.
  • Treat experimental models as research artifacts unless the card explicitly says they are production-promoted.
  • Keep EdgeGuard query-generation models behind schema validation, read-only execution controls, and bounded result inspection.
  • For LoRA adapters, load the adapter with the base model named in the model card.
  • Check license terms before using datasets or model outputs in commercial or redistributed work.