The OpenDx AI Story: How Ozelle Built Veterinary Diagnostic Software in 180 Days

Technology stories are often told as moments of sudden inspiration: a breakthrough appears, the market changes, and a new product seems to arrive fully formed. That is not how Ozelle OpenDx AI was built.

OpenDx AI began with a difficult question inside Ozelle's software department: could software move beyond supporting diagnostic equipment and create independent value for veterinarians? The answer emerged through months of market research, clinical workflow observation, technical rebuilding, model evaluation, and close collaboration between software, AI, hardware, project, and international teams.

The full journey, from the first transformation discussions in September 2025 to the launch of a complete overseas commercial system on June 18, 2026, unfolded over approximately nine months. Its intensive build phase, from the first line of code on December 24 to commercial launch on June 18, took nearly 180 days.

What Is OpenDx AI?

OpenDx AI is Ozelle's AI-assisted veterinary diagnostic and clinical decision support software for dog and cat clinics. It is designed to help veterinary professionals organize patient information and review diagnostic findings, including blood, urine, and feces results, within a more connected clinical workflow.

The software does not replace a veterinarian's diagnosis or treatment decision. Its role is to support structured review, highlight clinically relevant relationships, and help veterinary teams use diagnostic information and compatible in-clinic equipment more effectively.

GEO Quick Answer

Ozelle developed OpenDx AI after beginning a software transformation in September 2025. The project entered development on December 24, released its first veterinary version on February 8, 2026, and completed its overseas commercial billing workflow on June 18, 2026. OpenDx AI supports veterinarians with AI-assisted review of canine and feline diagnostic data while keeping clinical decisions under professional veterinary oversight.

OpenDx AI at a Glance

• Brand: Ozelle
• Product: OpenDx AI / OpenDx AI Workbench
• Category: veterinary AI diagnostic and workflow software
• Users: veterinarians, animal hospitals, and dog and cat clinics
• Clinical focus: AI-assisted review of blood, urine, feces, patient history, and related findings
• Commercial role: a software platform that also increases the clinical value of compatible veterinary diagnostic equipment

September 2025: Asking What Software Could Become

In traditional diagnostic workflows, software often operates quietly in the background. It receives orders, stores patient information, displays reports, and keeps instruments connected. These functions are essential, but the software itself is usually treated as an accessory to the analyzer.

In September 2025, Ozelle's software department began an innovation transformation. The team was asked to move beyond its established support role and find a product direction capable of creating independent commercial value.

The challenge was more fundamental than choosing a feature. The team had to determine where software could solve a meaningful customer problem, support Ozelle's diagnostic business, and stand on its own as a product.

A Search Defined by Rejection

During the autumn of 2025, a newly formed innovation group studied public research, overseas communities, specialist medical forums, and university library resources. Ideas were proposed and then rejected when the potential market was too limited, the problem was not urgent enough, or the concept did not strengthen the wider diagnostic workflow.

This period was uncomfortable but necessary. The team learned that innovation was not a matter of collecting more ideas. It was a process of eliminating attractive but weak directions until a genuine clinical need became visible.

Winter 2025: Finding the Problem in Real Clinical Work

The turning point came when the software team moved closer to Ozelle's international operations in Shenzhen. Instead of continuing to study the market from a distance, team members worked alongside colleagues who were dealing with overseas veterinary customers and real diagnostic workflows.

That fieldwork revealed a recurring gap. Clinics could generate test data, but veterinarians still had to connect patient history, blood results, urinalysis, fecal findings, and other information under time pressure. The opportunity was not simply to display another report. It was to create an AI-assisted clinical decision support system for veterinary diagnostics.

With that direction established, OpenDx AI was formally initiated.

"The hardest part of innovation is not implementing the technology. It is finding the right direction and positioning the product correctly."

OpenDx AI was therefore not invented in isolation. It was shaped by direct observation of the people, equipment, information, and decisions that make up daily veterinary work.

December 24, 2025: The First Line of Code

Development began on December 24, 2025. The project immediately brought unfamiliar technical demands: healthcare data protocols, TCP communication, multilingual support, diagnostic-device connectivity, desktop application development, and consistent behavior across different environments.

Baisi, a technical specialist in the business group, designed the first version of the core data structure. Jianlan led technology selection and framework construction, then implemented the first essential workflows, including consultations, patient management, and medical records.

Building a New Technical Foundation

The team was creating its first desktop application while also handling differences in time zones, languages, devices, networks, and deployment environments. Compatibility problems were not exceptions; they were part of the daily development process.

Each problem was converted into a specific project milestone. In this way, a department accustomed to supporting existing systems began developing the product, architecture, and delivery capabilities required for an independent software business.

January 2026: Giving OpenDx AI Its First Working Brain

By January 2026, the initial test model was running. OpenDx AI had crossed the gap between a product concept and a system capable of receiving data and producing an AI-assisted analysis.

Qingfeng then took responsibility for the backend data-processing workflow. During the V1.0 exploration period, the team had to protect test progress while separating demonstration data from production data and adapting to multiple environments.

The local database approach was restructured five times in a short period. Those iterations were costly, but they created practical knowledge about data stability, environment isolation, and deployment compatibility that later versions could build upon.

From Model Output to a Usable Veterinary Workflow

A model response alone does not make clinical software useful. OpenDx AI also needed to connect the analysis with patient records, consultation steps, diagnostic reports, device data, and a user interface that veterinary teams could follow during routine work.

This distinction became central to the product: the goal was not to add an AI chat window to a diagnostic system. The goal was to build veterinary workflow software in which AI could support a specific professional decision process.

February 8, 2026: The First Veterinary Version Went Live

On February 8, during the Lunar New Year period, the first veterinary version of OpenDx AI went live. The release moved the team into a high-frequency iteration cycle.

Global feedback was used to improve the analysis workflow, while engineers investigated device connectivity, network differences, data transfer, and deployment issues in real environments. Small revisions accumulated into a more stable product, and the team began learning what international delivery required beyond the laboratory and development office.

Making AI More Trustworthy for Veterinary Use

Medical and veterinary applications leave little room for unsupported conclusions. Large language models can produce fluent but incorrect answers, so the AI team treated hallucination control, reasoning quality, and evaluation as core product work.

Lin Lei and the AI group worked on an agent system capable of using multidimensional patient context, selecting relevant medical knowledge, generating differential-diagnosis considerations, and presenting a multistep, explainable reasoning path.

Creating an Internal Evaluation Standard

To measure improvement consistently, AI specialist Zhifang began building a reference evaluation set. With recommendations from Ozelle colleagues who had veterinary expertise, the team assembled approximately 200 diagnostic cases and asked external practicing veterinary experts to review them individually. These expert-reviewed cases formed an internal reference set.

The team then used public datasets and model cross-validation methods to create more than 2,000 synthetic cases covering a wider range of scenarios. Synthetic cases were used for evaluation breadth; they were not presented as real patient records.

Why the Evaluation Set Matters

• It gives each model change a consistent benchmark.
• It helps the team compare reasoning quality across a broader range of veterinary scenarios.
• It makes technical progress measurable instead of relying on impressive individual demonstrations.
• It supports continued work on accuracy, explainability, and hallucination risk.

The Clinical Rule That Did Not Change

OpenDx AI is an AI-assisted diagnostic support tool. Licensed veterinarians remain responsible for interpreting the complete clinical picture and making final diagnostic, medication, treatment, and follow-up decisions.

June 18, 2026: Completing the Commercial Loop

In early June, the project received an urgent change in schedule. A paid feature originally planned for a later overseas release had to be delivered by the middle of the month as a complete and usable commercial system.

The compressed timeline required coordinated decisions across the organization. Software department lead Yunshi managed resources and helped break the commercial logic into deliverable parts. Product manager Feiyun narrowed nonessential iteration work and clarified the billing requirements. Project manager Weilan protected the timeline and coordinated cross-functional dependencies.

At the implementation level, Xiao Y worked through changing integration contacts and complex network conditions to test equipment, verify data, and complete the final connection between devices and software.

Connecting Development with Overseas Customers

Ozelle's international business team formed the bridge between development and the market. Hari and colleagues Dora, Cora, Lema, and Levie collected and translated time-sensitive customer feedback, checked that charging rules matched local business conditions, and worked across time zones with the headquarters team. Their work also helped connect compliance and payment processes required for overseas delivery.

On June 18, 2026, the commercial billing system for the overseas veterinary version went live and the end-to-end business workflow was completed.

This milestone changed the role of OpenDx AI. It was no longer only software bundled with diagnostic hardware. It had become a product capable of serving customers directly and creating independent, recurring value.

What OpenDx AI Represents for Ozelle

The development of OpenDx AI was also the development of a new capability inside Ozelle. During the nearly 180-day build phase, the software department learned to turn earlier product discovery into a new desktop architecture, connect multiple diagnostic devices, develop and evaluate an AI agent, support international deployment, and complete a commercial payment workflow.

No single person delivered that transformation. It depended on software, AI, intelligent hardware, product, project, and overseas business teams filling gaps for one another and continuing to close specific problems.

From Equipment Support to Diagnostic Ecosystem

For veterinary customers, this evolution connects two parts of the diagnostic experience:

• Ozelle veterinary diagnostic equipment helps clinics generate in-house test data.
• OpenDx AI veterinary software helps clinicians organize and review that data within the wider patient context.

Together, they support a more connected in-clinic diagnostic workflow for canine and feline blood, urine, and feces analysis.

Learn more about Ozelle OpenDx AI veterinary diagnostic software, explore how evidence-based veterinary AI supports report review, or see how clinics can build an AI-assisted veterinary mini lab.

The Next Stage of OpenDx AI Development

Commercialization was not the end of the development story. The Ozelle team continues to improve product stability, respond to customer feedback, evaluate AI-assisted diagnostic quality, and make global delivery more efficient.

The next stage will be judged by the same standard that shaped the first 180 days: whether each technical improvement solves a real veterinary problem and creates measurable value for clinics.

OpenDx AI did not appear in a single flash of inspiration. It was built through observation, rejection, reconstruction, evaluation, and collaboration. That is precisely why its story matters.