How a startup team from Northwestern Polytechnical University with a valuation of over 1 billion successfully established a foothold in industrial scenarios

While many robotics manufacturers are still searching for viable real-world use cases to move from 0 to 1, this startup team originating from Northwestern Polytechnical University has already begun developing fourth-generation embodied intelligent industrial robots. 

Recently, Sixth Mirror Technology completed a Series B1 financing round of tens of millions of yuan, with investments from Kunlun Trust and Xi'an High-Tech Investment. The company's valuation has exceeded 1 billion yuan, and it has simultaneously launched its Series B+ financing round. While most AI enterprises are still grappling with confusion over "how to deploy technology practically," this AI firm rooted in Northwestern Polytechnical University has long established a solid foothold in two core scenarios: industrial quality inspection and workplace safety.

At its core, what Sixth Mirror builds is industrial AI: leveraging artificial intelligence to empower the transformation and upgrading of traditional manufacturing industries. And industrial AI is precisely positioned at the intersection of dual favorable trends from policy support and industrial development. From China's national "14th Five-Year Plan for Intelligent Manufacturing Development" to the "Action Plan for the Development of Intelligent Inspection Equipment Industry," from the introduction of "new quality productive forces" to the intensive rollout of local "AI + Manufacturing" special initiatives, policymakers are paving a fast track for the practical deployment of industrial AI.

Three Major Barriers Facing Industrial AI

The industrial AI track has never been an easy path to navigate.

Diametrically different from the consumer internet's characteristics of "massive data and standardized scenarios," the integration of AI with industrial production faces three inherent natural barriers:

First barrier: Extremely scarce small-sample data constraints. First, a critical distinction needs to be made: not all industrial scenarios require AI to "learn what constitutes a defect." For products like PCB boards and mirror-finished steel strips that feature highly consistent characteristics, AI only needs to learn "what a qualified product looks like" and can detect anomalies by identifying deviations from that standard.

However, in high-end industrial production environments, production lines typically operate with extremely high yield rates, making defective samples inherently rare. To enable AI to learn to identify "what a defect looks like," you first need access to a sufficient number of "defective samples" — which in itself creates a paradox. Traditional deep learning relies on massive volumes of data to drive model performance, but in industrial scenarios, data collection is difficult and costly, leaving model training stuck at a standstill.

Second barrier: Highly specialized industry knowledge requirements. Unlike the relatively standardized production processes and quality inspection standards in the 3C and lithium battery sectors, industries such as metallurgy, chemicals, and energy feature highly unique, non-interchangeable production workflows, equipment mechanisms, defect characteristics, and quality evaluation systems. There are no universal inspection algorithms or model logic that can work across all scenarios, nor standardized solutions that can be reused across different industries. An algorithm model optimized for steel metallurgy quality inspection cannot adapt to inner-wall corrosion and crack detection in petroleum pipelines; a technical system tailored for identifying sheet metal defects will completely fail when transferred to pipe and pressure vessel inspection scenarios.

This type of high-end industrial AI inspection is far more than simple algorithm R&D — it must be deeply integrated with core industry know-how. Accumulating specialized industry knowledge, on-site scenario experience, and defect judgment logic requires long-term, hands-on refinement in real production environments. These capabilities cannot be quickly replicated or rapidly deployed, creating an extremely high barrier to industry entry.

Third barrier: Full-link closed-loop capabilities from perception to execution. Currently, most industrial AI enterprises only excel at front-end visual algorithms. In general scenarios like 3C manufacturing and lithium battery production, solutions only need to complete defect perception and labeling to be deployed, without requiring involvement in subsequent handling processes. However, high-end industrial scenarios demand a complete full-link closed loop: not only accurate perception, identification, and decision-making, but also integration with on-site equipment, production workflows, and operation and maintenance systems to complete intelligent handling, status calibration, data review, and model iteration.

These scenarios feature strong operational isolation, high professional requirements, complex equipment interconnection, extremely low fault tolerance in production, and strict operation and maintenance specifications. Most enterprises lack practical on-site industrial operation experience and system integration capabilities, so they can only achieve basic perception functions and cannot support subsequent decision-making, handling, and review processes — completely breaking the value closed loop of "perception → judgment → handling → review → iteration." Precisely because full-link technology is highly complex, adaptation costs are high, and validation cycles are long, moving a product from 0 to 1 for practical deployment typically takes years, far longer than in general inspection scenarios.

Entering the Market in 2019 and Securing "National Team" Clients

As early as 2019, Sixth Mirror Technology began deploying embodied intelligence scenarios. China National Petroleum Corporation, HBIS Group, Taigang Group, Xinxing Ductile Iron Pipes, Shaanxi Coal and Chemical Industry Group — these "national team" enterprises within China's industrial system have been clients that Sixth Mirror has gradually served since 2019. Today, Sixth Mirror's product matrix covers a wide range of scenarios across the industrial sector. Its recent product launch event focused primarily on two practical deployment use cases for its embodied intelligent robots.

Launched Scenario Product: AI + Quality and Safety Assurance for Invisible Industrial Spaces

At the launch event, Liu Chuang, founder of Sixth Mirror Technology, demonstrated the three-layer technical architecture of Sixth Mirror's industrial robots: the bottom layer consists of AI toolkits, the middle layer is the "eyes and brain" combining hardware and software such as computing power systems and 2D+3D visual recognition systems, and the top layer is the embodied execution layer that directly serves end-user applications. This architecture fully showcases Sixth Mirror Technology's full-stack technical capabilities in the high-end industrial sector.

Scenario 1: Prismind L Pipeline Visual Quality Inspection Robot.

Oil and gas pipelines, chemical transmission pipelines, municipal water supply and drainage systems... these pipelines are often hidden in dark, narrow, high-risk unstructured extreme environments. Traditional inspection methods either cannot reach these spaces at all, or come with exorbitant costs and low efficiency. Sixth Mirror's Prismind L pipeline robot is designed to solve four core challenges: clear visibility, stable movement, accurate judgment, and comprehensive management.

Clear visibility — It penetrates dark, narrow, high-risk unstructured extreme environments, breaking through the detection limits of human vision and conventional methods, converting the previously invisible internal state of pipelines into perceivable, diagnosable, and decision-supporting data assets.

Stable movement — It can travel reliably through the complex, unstructured interiors of pipelines, adapting to confined spaces such as small and medium-diameter pipelines.

Accurate judgment — Equipped with a deeply customized vertical-domain large model built specifically for industrial quality inspection scenarios, it possesses "expert-level" industrial knowledge to deliver precise assessments of inspection data.

Comprehensive management — It forms a complete full-link closed loop from perception, identification, judgment, and handling to post-operation review, which is consolidated into an iterable, transferable, and scalable industrial embodied intelligence capability.

For industries like oil and gas and chemicals, Prismind L solves long-standing pain points of traditional inspection models: low coverage, delayed response, and high costs. It transforms pipelines from unobservable "black boxes" into visualized, data-driven transparent systems, enabling a paradigm shift from "reactive maintenance after failures" to "predictive maintenance." For frontline inspection and operation personnel, it replaces human workers in dark, narrow, high-risk environments, drastically reducing safety hazards. For enterprise decision-makers, inspection data is no longer a "one-time consumable" — it becomes a core asset that can be iterated and reused.

Prior to this launch, Sixth Mirror had already rolled out the Prismind series of quality inspection robots tailored for steel products of various shapes, including profiles, sheets, and wires. The release of Prismind L marks the completion of Sixth Mirror's industrial quality inspection product line coverage across all categories of the metallurgy industry — spanning profiles, sheets, pipes, wires, and other application scenarios. 

Scenario 2: "Eye-Brain" Embodied Intelligence System for Workplace Safety.

If the pipeline robot solves "quality" challenges, the workplace safety solution addresses issues that are literally "life-or-death."

In high-risk industries such as metallurgy, energy, and mining, workplace safety is far more than a slogan — it directly impacts human lives. Traditional workplace safety monitoring relies on cameras paired with manual oversight. But the reality is that the massive volumes of video streams generated by thousands upon thousands of cameras are impossible for human operators to process fully; and single algorithm models frequently suffer from false alarms and missed detections under extreme operating conditions such as heavy dust, low-light obstruction, and high-temperature radiation.

Developing a single algorithm is relatively straightforward, but building a solution that covers the full workflow SOP (Standard Operating Procedure) of complex industries is extraordinarily difficult.

Why is it so challenging? Because the SOP for workplace safety goes far beyond simple tasks like identifying hard hats and reflective vests. It involves specific, granular industrial production steps — for example, how to safely replace a flange. Replacing a flange is not a simple disassembly and assembly task: it is a high-risk, complex operation that requires going through procedures including work permit approval, energy isolation and lockout-tagout, medium evacuation and purging, and blind plate installation. During installation, workers must use torque wrenches to tighten bolts diagonally in stages to precisely control gasket compression. After the operation, strict pressure testing and leak detection are required to verify sealing integrity. Any oversight can lead to leaks or even catastrophic accidents. Consolidating SOPs for such complex industrial workflows requires extensive professional knowledge and practical experience — which is exactly the core difficulty of deploying complex algorithms in real-world industrial settings.

Sixth Mirror's solution is a full-link intelligent safety management system built around the "TianCe" Workplace Safety Vision-Language Large Model and the "XiYi" AI Safety Decision-Making Brain.

At the perception layer, it integrates multiple cameras, sensors, and robots to monitor environmental anomalies in real time and identify unsafe behaviors. At the management and control layer, it supports full-process monitoring of high-risk operations and compliance checks for work permits. At the emergency response layer, it provides knowledge Q&A and recommended handling procedures. At the reporting layer, it automatically generates reports that can be exported with one click. At the decision-making layer, it aggregates operational status across multiple production facilities to support real-time decision-making by senior management. The value of this solution extends to every stakeholder from frontline workers to executive leadership: frontline personnel receive real-time alerts to avoid hazards; safety managers reduce their workload while enabling automatic closed-loop management and compliance; senior executives gain real-time operational insights for decision-making. Most importantly, it offers lightweight SaaS services that reduce decision-making burdens for small and medium-sized enterprises, allowing them to access AI capabilities through low-cost subscription models. Following six steps — pre-configured templates, data integration, model fine-tuning, edge deployment, business configuration, and operational optimization — the solution delivers transferable, combinable, and scalable capabilities. This full-process closed loop from perception to handling and review is the core capability of Sixth Mirror's embodied intelligence technology.

"Slow from 0 to 1, Fast from 1 to 10": The 10-Year Development Logic of an Industrial AI Enterprise

"Industrial AI is the future direction of artificial intelligence development," Liu Chuang, founder of Sixth Mirror, stated during an interview. In 2014, Liu Chuang — then a junior at Northwestern Polytechnical University — co-founded Sixth Mirror with two classmates. From a campus startup to today's enterprise with a valuation exceeding 1 billion yuan, they have traveled this path for 12 years. Unlike consumer-grade AI that pursues broad applicability, industrial AI prioritizes depth: you must dig deeply and thoroughly into a single scenario before expanding horizontally. This is exactly Sixth Mirror's development philosophy: to cultivate deep expertise across different industrial sectors, then distill reusable industry methodologies from that experience.

"Industrial scenarios progress slowly from 0 to 1, but the improvement from 1 to 10 can happen very quickly. Because once you truly understand the pain points of an industry, accumulate sufficient on-site data, and refine a reusable standardized product, the marginal cost of horizontal replication drops dramatically." Today, Sixth Mirror has served over 500 industry clients and created 726 AI-empowered scenarios. From steel to chemicals, from energy to mining, from domestic markets to overseas operations — this "1 to 10" flywheel is accelerating.

Standing on the Shoulders of China's Manufacturing Industry, Bringing Industrial AI to the Global Stage

Beyond China's domestic market, feedback from overseas markets has also been extremely encouraging. Sixth Mirror's products are gradually expanding internationally, with established partnerships with clients in countries including the United Arab Emirates and Russia.

In March 2026, Sixth Mirror partnered with SICK — the global industry leader in industrial sensor and vision technology from Germany — to co-exhibit at Vision China in Shanghai. Liu Chuang recalled: "During discussions with overseas clients, one foreign partner offered this evaluation of our products: 'Your company's technology makes me feel like I'm looking into the future.'"

This is no exaggeration. In the global industrial AI race, Chinese enterprises, supported by the vigorous growth of China's manufacturing industry, its vast industrial market, and extensive client base, have already reached the global leading edge. Sixth Mirror's industrial embodied intelligence technology has grown precisely on the foundation of China's manufacturing sector — which boasts staggering steel output, exceptional chemical production capacity, and a massive network of energy infrastructure. In the future, China's manufacturing industry will gradually move to the center of the global stage, and industrial AI enterprises will be able to go even further as a result.

From a startup lab at Northwestern Polytechnical University to a leading industrial AI enterprise valued at over 1 billion yuan, Sixth Mirror's story is, in a sense, a microcosm of China's hard-tech startup ecosystem: it chases no fleeting trends, only solves real problems; it does not hype empty concepts, only delivers practical deployments. And the future of China's manufacturing industry precisely needs more of this kind of genuine, proven capability.