BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//sf26-agenda//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:SESS-1@sf26-agenda
DTSTAMP:20261020T151709Z
DTSTART:20261105T001500Z
DTEND:20261105T010000Z
SUMMARY:The Model Isn’t the Product: Building Streaming Evidence Layers f
 or Industrial AI
LOCATION:Room 2000
DESCRIPTION:Tony Lo\n\nMost AI demos assume clean data\, stable networks\, 
 and users who live inside dashboards. Factory floors do not work that way.
  This session shares how IntelFactor.ai is building an edge-first manufact
 uring intelligence system for real-time quality inspection\, where local d
 evices run computer vision inference\, operators receive PASS / REVIEW / F
 AIL decisions on the line\, and the cloud receives structured inspection r
 ecords\, evidence\, review actions\, and telemetry for analytics and AI-as
 sisted root-cause analysis. The core theme is simple: in industrial AI\, t
 he model is not the product. The product is the evidence layer that turns 
 physical events into trustworthy\, reviewable\, and actionable data stream
 s. We will walk through the technical architecture behind this system\, in
 cluding NVIDIA Jetson edge inference\, camera-based inspection\, local per
 sistence\, evidence capture\, HTTP-based inspection sync\, Kafka-style eve
 nt modeling\, telemetry streams\, and cloud workflows for review and analy
 sis. The talk will also cover where streaming fits\, where it does not\, a
 nd why forcing every workflow into a topic can turn a practical system int
 o a distributed systems museum exhibit. Attendees will learn how to design
  production AI systems that connect edge inference\, structured events\, h
 uman-in-the-loop review\, and cloud analytics. Key takeaways include how t
 o model inspections as events\, how to separate continuous telemetry from 
 discrete inspection records\, how to preserve traceability\, and how to bu
 ild systems that continue operating when networks\, operators\, and realit
 y behave badly\, as they always do. This aligns well with Current’s sugg
 ested themes around AI in production\, real-world applications\, operation
 al excellence\, and edge-based streaming architectures.
URL:https://sf26.rmoff.net/s/the-model-isnt-the-product-building-streaming-
 evidence-1
END:VEVENT
END:VCALENDAR