Tangyu Zhiyi
A diabetes Q&A module within a broader medical knowledge augmentation research effort. As first author, worked on the frontend, backend, LoRA fine-tuning and performance testing.
PROJECTS / IDEAS INTO PRACTICE
From AI research to practical tools and creative applications.
Explore the ideas, technologies, and public links behind each project.
A diabetes Q&A module within a broader medical knowledge augmentation research effort. As first author, worked on the frontend, backend, LoRA fine-tuning and performance testing.
A desktop tool for writing image captions and labels, bringing image browsing, data organization and annotation into one workspace.
Maintain a local 3D printing management application for filament inventory, historical records, printer status and device monitoring.
Organize software copyright application materials with syntax highlighting, document layout and PDF export for reviewable source-code documents.
Led a document understanding and multi-source data fusion project, exploring LLM tool calling and tasks involving Word and Excel documents.
Led research on medical document classification using image and text information, with DiT, BERT and late fusion.
Online service ending soon
As technical lead, worked on CT/DICOM image processing, YOLOv8 detection and model validation.
Explore knowledge graph construction and retrieval with multi-source traditional Chinese medicine data, organizing distributed knowledge for information search.
An engineering experiment in LLM conversations and plan recommendations for chronic disease scenarios, connecting a web interface, backend APIs and local storage.
Simulate face-based access control with camera and image recognition plus access logs to exercise application interactions and processing flows.
A multi-role application for herbal decoction prescriptions and task workflows, organizing prescription records, task handoffs and role-based collaboration.
A Vue-based personal website with build-time prerendering and Python content validation, bringing together personal information, projects, and useful resources.
Extract paper PDFs locally, use an LLM to organize the main ideas and evidence, and generate HTML/A1 posters for offline viewing and printing.
A local web prototype for understanding health reports, exploring structured extraction and checks against original evidence to organize information and its sources.
Search papers by keyword, organize journal information and export results as JSON, CSV or Excel for later review and processing.
Aggregate Claude Code usage and estimated costs across devices, view statistics by project and model, and organize collected records in a dashboard.
A demonstration of purchasing, sales and inventory workflows, with role permissions, task approvals and statistics pages to illustrate business processes.
A Vue music player interface exploring track presentation, playback controls and component interactions as part of the listening experience.
A personal blog application separate from this homepage, organizing reading and publishing around post editing, categories, tags and content management.
Online service ending soon
A personalized learning multi-agent prototype organizing learner profiles, resource generation and tutoring workflows.
Practice data structures, algorithms, object-oriented programming and computer science foundations with C++, Java and Python, collected as study code and notes.
The work keeps growing. So does the curiosity.
Explore more on GitHub(opens in a new tab)