DingTalk Rebuilds AI Ecosystem: Platform Trains Gynecology-Specialized Large Model with Over 90% Accuracy

Recently, the "Doukou Gynecology Large Model" developed by Yisheng Jiankang (Hangzhou) Life Science & Technology Co., Ltd. achieved an accuracy rate of 90.2% in professional testing. This marks the first high-accuracy, highly practical domain-specific large model successfully trained on the DingTalk AI platform. After deployment, the model will effectively alleviate the shortage of specialized gynecologists and help more female users manage their health.

DingTalk Empowers Industry-Specific Large Models

The successful deployment of the gynecology large model signifies that DingTalk's ecosystem has expanded beyond traditional SaaS, service provider, consulting, and delivery ecosystems to include AI entrepreneurs. As general-purpose AI large models become increasingly widespread, industries across the board are striving to deeply integrate large models into their specific business scenarios, building industry or domain-specific models to achieve cost reduction and efficiency gains in operations. To address challenges such as complex data engineering, high technical requirements, and data security during proprietary model training, DingTalk has established a comprehensive support system for enterprise and industry-specific large model development. This system assists companies throughout the entire process—from data annotation and model training to inference and deployment—enabling them to harness the productivity transformation brought by AI.

At 9:21 PM on June 30, good news came from Yisheng Jiankang’s office on the 5th floor of Hangzhou Lv Feng Center. Their large model, Doukou Gynecology Large Model, trained via DingTalk’s enterprise-exclusive AI platform, had just completed 100 professional gynecological test questions. The results showed that the model achieved a 90.2% diagnostic accuracy rate for six major gynecological symptoms, including menstrual disorders, abnormal bleeding, unusual vaginal discharge, lower abdominal pain, and pelvic masses, demonstrating significant practical value.

"The Doukou Gynecology Large Model acts like an AI gynecologist, offering pre-diagnosis and health management services to more women," said Wang Qiangyu, founder of Yisheng Jiankang. "Users simply select their symptoms in the chat interface of the Qinmi Doctor app and receive professional self-diagnosis results, including primary diagnosis, potential secondary diagnoses, recommended tests, treatment suggestions, and precautionary advice."

Compared to the average 30-minute wait time for traditional online consultations, Qinmi Doctor generates professional recommendations within seconds, helping women quickly determine whether urgent medical attention is needed—especially beneficial for working women and users in remote areas.

Yisheng Jiankang is a life science company specializing in women's precision testing and health services. Its founding team primarily comes from well-known internet companies, obstetrics and gynecology medical institutions, and biopharmaceutical firms. Based on technology trends and industry insights, Wang Qiangyu’s team believes that developing AI doctors through training gynecology-specific large models will effectively address the shortage of specialized gynecologists and healthcare services, creating substantial industrial and societal value for both medical aesthetics institutions and female users.

Just like cultivating real gynecological experts, a highly specialized "AI gynecologist" cannot be easily trained using generic large models. Since initiating the development of the Doukou Gynecology Large Model, Yisheng Jiankang’s team used open-source large models as a foundation and, through industry-specific data training, achieved a diagnostic accuracy of approximately 77.1% in the first version. "While 77.1% meets basic industry standards, for medical AI directly impacting health and safety, further breakthroughs are necessary—to shift from 'general knowledge coverage' to 'specialized expertise in vertical domains,'" said Wang Qiangyu.

After hitting a performance bottleneck, Yisheng Jiankang migrated the Doukou Gynecology Large Model training platform to DingTalk’s enterprise-exclusive AI platform. With DingTalk’s support, the team made comprehensive adjustments in data processing, increased computing power, and model optimization. Within one month, both parties raised the model’s diagnostic accuracy from 77.1% to 90.2%.

The significant enhancement in the Doukou Gynecology Large Model’s capabilities will strongly support Yisheng Jiankang’s rapid business growth. Moving forward, the company plans to further improve performance and accuracy, integrating updates into its flagship product, Qinmi Doctor. Through AI doctor agents, it aims to serve even more users. "Through this practice, we envision that in the future, vertical models for other fields—not just gynecology but also dermatology—can be developed and integrated into daily life, enabling ordinary people to receive preliminary health guidance comparable to that of professional medical institutions—all from home," said Wang Qiangyu.

"Improving a domain-specific large model’s accuracy from 77.1 to 90 is a major leap," said Zhu Hong, CTO of DingTalk. "It’s like transforming a generalist who knows a little about everything into a specialist-level expert in a short period. This involves effective preprocessing of data with strong security, efficient allocation of computing resources, construction of model evaluation mechanisms, and fine-tuning of training algorithms and model parameters—demonstrating DingTalk’s full-chain collaboration capability with industry partners in building domain-specific large models."

The Doukou Gynecology Large Model is the first vertical-specific large model developed with DingTalk’s assistance. Similar industry-specific large models and professional AI applications represent the next wave of AI adoption. As general-purpose AI models like Qwen, DeepSeek, and GPT evolve into foundational infrastructure, many enterprises can already access relatively standardized AI services through knowledge base construction. However, due to differences in industry-specific knowledge, operational scenarios, and workflows, businesses still lack practical experience and clear pathways when applying large model technologies to solve specific professional problems.

"Through collaboration with numerous partners, we’ve found that building, deploying, and applying proprietary large models presents challenges right from the start—uncertainty about where to begin and no guarantee of outcomes," said Zhu Hong. "Training and deploying large models in the cloud also introduces significant risks related to data flow."

To better assist enterprises across industries in training and deploying their own professional or proprietary large models, DingTalk has built a comprehensive support system for enterprise and industry-specific large model development. On DingTalk, companies and partners gain end-to-end platform and service support—including data collection, cleaning, annotation, base model selection, model training, performance evaluation, model tuning, and engineering deployment—enabling more efficient development of industry-specific large models and smoother implementation of AI applications. DingTalk also offers consulting services for industry AI solutions and large model strategies, along with AI talent training and certification programs, ensuring successful AI adoption for enterprises.

Underlying this effort is a restructuring of DingTalk’s ecosystem—from traditional SaaS, service provider, consulting, and delivery ecosystems to a new phase centered on aggregating AI entrepreneurs via the DingTalk platform. The Doukou Gynecology Large Model is just the beginning. For vertical-industry partners, DingTalk will leverage its open platform to help data-rich collaborators and developers build industry-specific large models and AI agents from scratch, then deliver these solutions through DingTalk’s app marketplace to meet the intelligent transformation needs of small and medium-sized enterprises across various sectors.

In the future, developers from all industries will be able to complete the full-cycle development of AI products on the DingTalk platform and achieve commercial closure, working alongside DingTalk to drive AI transformation across countless industries.

We dedicated to serving clients with professional DingTalk solutions. If you'd like to learn more about DingTalk platform applications, feel free to contact our online customer service or email at This email address is being protected from spambots. You need JavaScript enabled to view it.. With a skilled development and operations team and extensive market experience, we’re ready to deliver expert DingTalk services and solutions tailored to your needs!