Foxconn Boosts Production Throughput With DeepHow Live SOP Verification Powered By NVIDIA

Powered by NVIDIA Cosmos and NVIDIA Metropolis VSS Blueprint, DeepHow’s Live SOP Verification detects procedural drift, improves first pass yield, and reduces rework within Foxconn’s manufacturing center.

ROYAL OAK, Mich.--(BUSINESS WIRE)--DeepHow today announced an advancement in live SOP verification that brings physical AI into Foxconn’s manufacturing center, enabling teams to identify procedural drift and drive quality and productivity as work is performed.

DeepHow’s Live SOP Verification agent uses NVIDIA Cosmos and NVIDIA Metropolis Blueprint Video Search and Summarization (VSS) to give manufacturing, quality, and supply chain teams greater visibility into how standard work is executed in complex production environments.

As demand for accelerated computing infrastructure continues to grow, manufacturing consistency, throughput, quality, and supply chain execution are increasingly critical. Within Foxconn’s manufacturing plant, DeepHow’s Live SOP Verification is designed to support more consistent process execution by comparing real-world activity against approved procedures and mitigating deviations that can impact first pass yield, rework, and operational efficiency.

For years, manufacturers have relied on manual observation, periodic audits, and end-of-line inspection to understand whether work was performed correctly. Those methods often identify defects after they have already moved downstream, but rarely reveal the execution issues that caused them. Powered by the NVIDIA VSS blueprint, manufacturers can now reason over live video streams, work sequences, and operational context to guide workers in real-time and improve quality at the source.

DeepHow’s Live SOP Verification uses NVIDIA Cosmos to interpret complex human activity and work sequences in context, while NVIDIA Metropolis VSS enables video search, summarization, and analysis across operational environments. Together, these technologies help compare real-world execution against approved SOPs and surface actionable insights for supervisors, engineers, quality teams, and continuous improvement leaders.

This represents a major step forward for manufacturing AI. Traditional computer vision systems could identify objects and attributes, but they struggled to understand actions and procedural sequence. NVIDIA Cosmos vision language model and NVIDIA VSS Blueprint make it possible to reason over work as it happens, creating a foundation for live SOP understanding and process optimization at scale. DeepHow brings this capability to the factory floor, with its Live SOP Verification solution.

Within Foxconn’s manufacturing facility, DeepHow Live SOP Verification delivers:

• 3% improvement in First Pass Yield by providing real-time guidance
• 99% task-level accuracy in micro-action understanding of critical SOP steps
• Greater SOP adherence across complex, high-value production workflows, with every cycle analyzed rather than sampled
• Faster root cause analysis powered by powerful video analytics that pinpoint the source of process variation
• Continuous improvement at scale by turning frontline execution data into measurable operational intelligence across shifts, lines, and facilities

The engagement supports NVIDIA’s broader supply chain excellence initiatives by applying AI to the execution layer of operations to improve quality, productivity, and delivery performance. By connecting approved SOPs with real-time assistance and AI-driven insight, NVIDIA and DeepHow are helping close the loop between process standards, real-time execution, and operational outcomes. DeepHow is a member of NVIDIA Inception.

About DeepHow

DeepHow is a physical AI company for operational excellence, enabling manufacturers to capture knowledge, verify work, and optimize frontline execution at scale. Used by more than 100 customers across 1,500 locations in 28 countries, DeepHow serves manufacturers across industrial manufacturing, electronics, pharma, utilities, and food and beverage. Learn more at deephow.com.