News Center

位置：

Honors Announcement

Group News

Media Reports

Hebei Daily | “AI Empowers Thousands of Industries: Firsthand Stories | Visual AI Acts as a ‘Supervisor,’ While Algorithms Serve as ‘Strategists’ in Drug Screening”

Category:

Media Reports

发布时间：2025-05-22

【概要描述】 Recently, at Workshop 312 of Shijiazhuang Sihua Pharmaceutical Group, the upright soft-packaging production line for large-volume injectables has been operating at a rate of 400 vials per minute. Vial after vial of injectable solution “runs” into the “darkroom,” where strobe lights flash rapidly.

Hebei Daily | “AI Empowers Thousands of Industries: Firsthand Stories | Visual AI Acts as a ‘Supervisor,’ While Algorithms Serve as ‘Strategists’ in Drug Screening”

Recently, at Workshop 312 of Shijiazhuang Sihua Pharmaceutical Group, the upright soft-packaging line for large-volume injectables has been operating at a rate of 400 vials per minute. Vial after vial of injectable solution “runs” into the “darkroom,” where strobe lights flash rapidly.

Why take photos of injection solutions? In fact, this is part of the visual inspection process, with the “darkroom” being an AI-powered inspection machine.

Each vial of injectable solution undergoes a series of rigorous manufacturing steps, including formulation, filling and sealing, sterilization, visual inspection under lamp, and packaging. Among these, visual inspection under lamp serves as the “safety guardian” of the production line, tasked with detecting foreign particles and identifying any defects in the product’s appearance.

“In the past, lamp inspection relied entirely on the naked eye of a large number of experienced inspectors, which was not only inefficient and labor-intensive but also resulted in inconsistent inspection quality,” said Wei Peng, Deputy General Manager of the Second Manufacturing Center at Shisi Pharmaceutical.

Today, after injection vials are scanned by AI-based visual inspection machines, any nonconforming products are automatically rejected, with an inspection accuracy rate exceeding 99.9%. A small team of inspectors only needs to conduct random sampling and verification 2 meters from the exit of the “darkroom.”

How was this monumental transformation achieved?

The answer lies in the “keen eye” that Shisi Pharmaceutical has introduced—visual AI technology. By deeply integrating AI into the lamp inspection process, Shisi Pharmaceutical employs AI-powered lamp inspection machines to intelligently identify and inspect injectable solutions, significantly boosting production efficiency.

Training a qualified lamp-inspection technician is no easy task; it requires not only formal training and examinations, as well as hands-on skill practice, but also mentorship from experienced senior technicians, all of which demand substantial time and effort.

Today, by feeding domain expertise and vast amounts of data into visual AI models and subjecting them to iterative training, it is possible to rapidly develop AI models that meet inspection standards. In practical applications, visual AI technology leverages the cameras of automated visual inspection systems to continuously accumulate experience and engage in deep learning, thereby steadily improving detection accuracy.

“Back then, we didn’t have AI-powered lamp inspection machines. We worked in a darkroom, using white paper as the background and shining a powerful light on each bottle one by one for inspection,” recalled Bian Yan, a lamp inspector. “The workload was extremely demanding at the time, relying entirely on our eyes to make quick judgments. Prolonged work easily led to visual fatigue, making it difficult to maintain sustained high concentration.”

Now, with the support of visual AI technology, the working model of lamp inspection operators has undergone a fundamental transformation: they only need to conduct random inspections on qualified products.

“With the same production capacity, what once required 100 visual-inspection workers now needs only 20,” said Wei Peng. “In the past, manual visual inspection allowed each worker to inspect just over a dozen bottles per minute; today, with the support of computer-vision AI technology—and supplemented by a small number of inspectors for spot checks—we can achieve a throughput of up to 400 bottles per minute, making the entire production process much more seamless.”

Underpinned by digitalization and automation, Workshop 312 at Shisiyao has achieved seamless alignment between production and inspection speeds, thanks to the support of visual AI technology. Currently, the workshop’s daily output has surpassed 2 million bottles, demonstrating the substantial productivity gains enabled by AI.

If visual AI technology has equipped Shisiyao’s production lines with “keen eyes that see through fire,” then in the field of drug discovery, AI serves as the “superbrain” for precise insight, fundamentally reshaping the underlying logic of innovative drug R&D.

At the Shisi Pharmaceutical Research Institute, several quietly operating computers have become “super weapons” on the research battlefield. Researchers input commands, and the data on the computer screens continuously updates—this is the AI-assisted drug-screening platform automatically generating the data needed for the discovery and development of innovative drugs.

“Our innovative drug for the treatment of chronic kidney disease was developed using an AI-assisted drug-screening platform,” said Li Yuan, Deputy Director of the Innovation Department at the Shisi Pharmaceutical Research Institute. She explained that the platform not only elucidated the druggability of the target but also conducted intelligent compound screening and optimization, thereby facilitating the smooth progression of this drug into the PCC (preclinical candidate compound) validation stage and helping the team avoid many detours.

The AI-assisted drug-screening platform is an in-house AI system independently developed in 2022 by Li Yuan and her team as they integrated AI technologies into the drug-development process. Today, this platform has been applied across all facets of innovative drug R&D, serving as the “superbrain” of the Shisi Pharmaceutical Research Institute.

The research and development of innovative drugs is a highly challenging endeavor with an extremely high failure rate.

“In the past, developing a new drug was like navigating through thick fog; it once took our team eight years to bring a single candidate to market. During that time, we repeatedly faced uncertainty about our direction and even lost confidence,” said Li Yuan. She explained that in the early stages of traditional drug discovery—specifically, the initial screening phase—researchers could rely only on their experience to sift through vast libraries of compounds, searching for those with the potential to be developed into medicines. This process involved countless rounds of screening and experimentation, akin to looking for a needle in a haystack, often requiring four to five years and costing tens of millions of yuan.

With the AI-assisted drug-screening platform, the preclinical development of this small-molecule targeted inhibitor for chronic kidney disease—from target identification to compound screening and optimization—was completed in just one and a half years, while R&D costs were reduced to RMB 5 million, representing a dual breakthrough in both efficiency and cost. “More importantly,” Li Yuan said, “it has restored the team’s confidence in exploring the unknown.”

Currently, AI-assisted drug-screening platforms have already begun to deliver promising results in the discovery and development of innovative therapeutics for multiple disease areas, including epilepsy, diabetes, and kidney disease. Looking ahead, Li Yuan stated: “We will continue to upgrade our AI models and databases to build a more precise predictive system. At the same time, we will extend the application of AI technology into the clinical-trial phase, thereby establishing an intelligent end-to-end pipeline that seamlessly bridges the gap between the laboratory and the clinic.”

At Shisi Pharmaceutical, AI technology has been integrated across all facets of business operations. The intelligent production–supply–sales collaboration system addresses the pain point of the tedious and complex task of production scheduling; the manufacturing execution system enables the digitalization of batch records, moving them from paper-based to electronic format; and the order management and logistics transportation systems together form a closed-loop order fulfillment process. These AI-enabled modules are seamlessly interconnected, like precisely meshed gears, collectively propelling the company’s transformation toward intelligent operations.

Each application of AI technology is like a piece of a puzzle, gradually assembling into a comprehensive intelligent AI blueprint for Shisi Pharmaceutical. Leveraging the power of AI, Shisi Pharmaceutical is forging a “smart fast track” for high-quality development.

Originally published in the Hebei Daily on May 21.

Keywords:

Shijiazhuang Daily | Reading the Data, Seeing the Development | In the First Half of the Year, Shijiazhuang’s Value Added by Industries Above Designated Size Recorded Double-Digit Year-on-Year Growth—What Transformative Breakthroughs Do These Striking Figures Reflect?

Shijiazhuang Daily | “Front-Row Insights into Science and Innovation Platforms: Hebei Provincial Research Institute of Technology for Sterile Chemical Pharmaceutical Preparations—An Innovation Engine Accelerating Product Upgrades and Iterations”