1. Designed and implemented a full-stack medical insurance review platform with React + TypeScript frontend and FastAPI + MySQL backend, with Apache Iceberg tables on MinIO as the analytical store for audit logs and model features, enabling patients and insurance bureaus to process applications seamlessly.

2. Built batch auditing, real-time dashboards, and Excel export features to support large-scale claim processing, backed by Apache Spark pipelines that write aggregates to Apache Iceberg (on MinIO) for low-latency reads.

3. Integrated DotsOCR for multimodal document parsing (PDF/images) and Qwen3-30B LLM via vLLM for intelligent claim pre-review, backed by policy RAG to ground LLM reasoning in insurer policies, achieving structured JSON outputs and automated rule-based validation.

4. Designed retrieval-grounded, citation-required prompt templates with strict JSON-only output schemas and repair prompts for schema drift; tuned approve/deny/request-info and added few-shot adapters for new policies.

5. Engineered JWT authentication and RBAC, enforcing hospital-level data isolation and least-privilege access.

1. Designed and built a course management platform that allowed students and teachers to efficiently browse, register, and manage the course schedules through a user-friendly interface.

2. The system served over 1800 users per semester (800 DAU peak) and is handed over to co-founders for maintenance, aiming to serve tens of thousands of users across more than a dozen schools in future.

3. Developed responsive web and mobile apps using WeChat Developer Tools and Django REST Framework, implementing JWT-secured APIs, MySQL data storage, and smooth cross-device user experience.

Visited large-scale industrial projects, gaining insights into practical user interface design in the industrial sector. Actively participated in optimizing specific UI design and implementation for operational interfaces.

Experienced in Agile methodologies, Clean Code principles, and Test-Driven Development (TDD) with practical exposure to PON and access network technologies. Led key function implementations under the main window and implemented JUnit tests in Eclipse to drive development processes.

1. Developed a Web Application which demonstrates data visualization for driving behaviors, speeding, and mileage usage after integrating ~500G real business data from Shanghai Steel Union.

2. Analyzed and transformed data using Clustering Algorithms including Cohesive Hierarchy Method and K-means.

3. Using Vue, D3.js to draw bubble charts, river charts, and calendar heat maps based on the analyzed data.