Zhixian Zhang reports Chinese software copyright for decision support system

Zhixian Zhang said she obtained a Chinese software copyright registration on Oct. 9 for a data-driven decision support system tied to her graduate work in management and analytics at New York University. The filing sits alongside a broader portfolio that includes AI systems research, a lakehouse data architecture, a conversational agent and a creator recommendation tool.

Why it matters: - Zhang’s registration shows her work spans both AI infrastructure and decision-support software. - The portfolio connects research on system performance, data governance and application design into one technical profile. - The projects point to tools that can help organizations manage AI workloads, structure data and make planning decisions.

What happened: - Zhixian Zhang announced a Chinese software copyright registration dated Oct. 9 for “Data-Driven End-to-End Digital Management and Intelligent Decision-Making System V1.0.” - The registration number is 2026SR1008779. - Zhang is listed as the original copyright holder. - Zhang is a graduate student pursuing a Master of Science in Management and Analytics at New York University.

The details: - Zhang’s broader research and development work focuses on two engineering tasks: managing the resources needed to operate AI services and organizing data so users can evaluate information and make decisions. - Her publicly documented application work includes a conversational agent and a creator recommendation tool. - In AI systems research, Zhang studies how changing workloads affect response times and computing requirements. - Her work considers routing, processing configuration and resource allocation as demand changes. - Those operational choices affect whether an AI service can meet response-time requirements while controlling resource use. - One example is RT-TOP, a framework described in her conference chapter “Research on AI System Performance Modeling and Online Optimization Methods Based on Runtime Telemetry Data.” - Springer lists the chapter’s first online publication date as Sept. 2, 2026. - RT-TOP uses OpenTelemetry data, including traces, metrics and logs, to model performance during operation. - The framework combines that model with an optimization method that adjusts processing, routing and resource allocation within defined safety constraints. - The approach treats performance tuning as an ongoing feedback process based on measurements from a running service. - The chapter abstract reports up to 17.8% lower P99 latency and about 9.6% lower energy consumption per 1,000 requests while keeping the same service-level agreement in experiments using public Azure and Meta microservice traces. - P99 latency measures response time at the 99th percentile. - Those figures describe the study’s experiments and do not establish performance gains for the registered software or for deployments at the companies whose public traces were used. - Zhang’s enterprise data platform research addresses the data feeding those services. - In a separate 2026 paper in the International Journal of Big Data Intelligent Technology, Zhang proposed a lakehouse-based architecture that combines data contracts, rules expressed as code and version-controlled model deployment. - Data contracts specify how data should be structured and validated. - Traceable metadata records where data came from and how it was processed. - The design links governance with software engineering so changes to data and models can be tracked as an AI pipeline evolves. - Zhang’s proposed release process includes contract tests for data structure and quality, followed by a limited rollout. - If a release fails, the design calls for restoring the previous version. - Zhang also developed an AI Conversational Agent as a separate application project. - The application connects a language model with stored knowledge and conversational context. - Its public documentation describes a modular design built with FastAPI and LangChain. - The application uses Redis for conversational memory and Qdrant for vector-based retrieval. - The design also includes integrations for web search and speech processing. - Zhang organized the application around components for model access, information retrieval, memory and response formatting. - The workflow makes the handoff between stored context, retrieved material and generated answers visible in the documented design. - Retrieval pulls relevant material from a knowledge base to support the model’s response. - Conversational memory retains information needed across exchanges. - External search provides another source of information. - The accompanying browser demonstration is labeled an interactive prototype and defaults to an offline mode. - The repository documents the fuller application design. - A second application focuses on decision support in a marketing workflow. - In September 2026, Zhang worked on an AI-assisted creator discovery and recommendation project. - The Creator2 repository documents a process that starts with a campaign brief. - The project uses the YouTube Data API to identify candidates and gather channel information. - It combines quantitative indicators with Gemini-assisted assessments to rank creators. - The application evaluates engagement, publishing consistency and audience fit. - It then applies filters and a budget constraint to recommend a selection. - The documentation distinguishes metrics derived from platform data from assessments inferred by AI. - That separation shows which parts of a recommendation come from observed indicators and which come from model interpretation. - Budget calculations use estimated costs, so the output is planning support rather than a record of campaign results. - Users can adjust scoring weights to reflect campaign priorities. - The tool also records reasons for excluding candidates, including audience mismatch or inactivity. - It compares selections under different budget assumptions.

Between the lines: - Zhang’s work suggests a consistent theme: making AI systems more measurable, governable and explainable. - The mix of operational research and application projects points to interest in the full stack, from infrastructure through user-facing decision tools. - The documentation emphasis on metrics, filters, rollback and source distinctions suggests a focus on auditability as much as performance.

What's next: - Zhang’s registered software may support further development or broader use as her research and application work evolves. - Her publication and project pipeline suggests additional work could emerge around AI operations, data governance and recommendation systems. - The documentation and prototype structure also leave room for continued iteration on both the conversational agent and creator-selection tools.

The bottom line: - Zhang’s new copyright registration adds a formal marker to a technical portfolio that links AI performance tuning, data architecture and decision-support software.