InnerOS
Governed AI Operations on AMD
VERIFIED EVIDENCE READY
AI reasoned · human governed · evidence captured
AMD Developer Hackathon · ACT III

Governed AI Operations on AMD

Reason. Act. Prove.

Gemma LIVE VERIFIED Evolus LIVE VERIFIED Evidence REPLAYABLE
AI-Native Incident Resolution Pipeline
⚠
SOURCE

Bellini Incident

Optical flap + camera telemetry loss

DETECTED
◉
AMD TOKEN FACTORY

Gemma on AMD

Gemma-4-31B-IT · ROCm

READY
⌘
BUSINESS AGENT

Evolus

Workflow + Knowledge Library

READY
✓
INNEROS

Evidence

Verify · seal · replay

READY
AI REASONING · GEMMA ON AMD

Operational diagnosis

READY
Select Run Live Demo for a fresh execution, or replay the verified ACT III evidence receipt.
ModelGemma-4-31B-IT
RuntimeAMD ROCm
Latency—
Tokens—
INCIDENT PLAN

AI-directed response

NO RUN
  1. 1
    Awaiting incident analysis

    Gemma will produce a structured IncidentPlan.

EVOLUS WORK ORDER

Not created

READY
WorkflowKnowledge Library
Waiting for governed remediation workflow
Google DeepMind GemmaLIVE VERIFIEDGemma-4-31B-IT on AMD Token Factory
Evolus Business AgentsLIVE VERIFIEDWorkflow + Knowledge Library over MCP
InnerOS EvidenceREPLAYABLESHA-256 execution bundle + truth labels
LOCAL AMD EXECUTION · MEASURED

Radeon AI PRO R9700

Qwen3-Coder-30B-A3B-Instruct-AWQ · vLLM · ROCm

0.2622 s N=20 p500.2638 s p9522.85 tok/s
VOICE TO GOVERNED WORK

Inner Oz → InnerOS → GitHub + Ops

Signed simulator proof created the canonical demo repository and governed task. Physical Echo confirmation and assigned coding completion remain pending.

Inspect the voice-created repository ↗
COORDINATION CROSS-TEST · VERIFIED

6 external checks · 8 swarm lanes

All checks passed with unique nonces. This is cross-test coverage, not a claim of eight simultaneously active agents.