WARPSENSE│AI INFERENCE ENGINE
SYSTEM ONLINE

ON-DEVICE WELDING AI

FunctionGemma 270M · Cactus Runtime · Gemini Cloud Fallback

MODEL
270M
params
ON-DEVICE LATENCY
<100
ms
ROUTING
SMART
interrogative gate
OFFLINE
YES
zero network
DEMO SCENARIOS
01ON-DEVICEOn-Device Speed
"voltage is 28V, range 18-24"
02ON-DEVICEOn-Device Speed
"angle 52 degrees, max 45"
03ON-DEVICECloud Coaching
"session WS-042 scored 41, why?"
04ON-DEVICEAgentic Chaining
"Analyze session WS-042 — check score, then parameters, then flag issues"
05ON-DEVICEGraceful Escalation
"wire_feed is 12, range 10-15"
OUTPUT
SELECT A SCENARIO OR RUN ALL
LIVE QUERY — POST /api/ai/analyze
ON-DEVICE
FunctionGemma 270M via Cactus runtime. <100ms. Sensor data never transmitted.
ROUTING
Interrogative gate: why/explain → cloud. Measurement queries → on-device.
ESCALATION
Enum validation failure triggers automatic on-device→cloud escalation.
OFFLINE
Zero network fallback. Graceful degradation for shop floor connectivity.