About Polyrek Research Hub
Project Positioning
Polyrek is an independent research project exploring new ways to structure technical knowledge from published patents.
Rather than treating patents only as documents to be searched, the project investigates whether technical disclosures can be decomposed into smaller, reusable technology concepts and reorganized into higher-level knowledge structures.
The long-term goal is to develop a more systematic way to observe technology evolution, identify relationships between technical concepts, and support technology intelligence and R&D decision-making.
Who am I
My name is David Ku, and I currently work as an in-house patent engineer at a technology company.
Throughout my 20 years of experience in the patent field, whether in patent portfolio and prosecution management, patent drafting, or prior-art search and analysis, I have consistently sought to bring data-driven approaches into my work. Before the widespread adoption of large language models, this kind of effort was often described simply as “digitalization.”
I see 2026 as a turning point for LLMs—from models primarily used for content generation to systems capable of reasoning, tool use, and increasingly autonomous execution. This shift also became the catalyst for my research into AI-assisted knowledge representation.
Intended Outputs
- Research questions and validation plans
- Experiment records and decision logs
- Analytical notes and methodological framing
- Domain-specific showcases for technology intelligence use cases
Collaboration Interests
Polyrek welcomes collaboration with researchers, domain experts, and software developers interested in technology intelligence and AI-assisted knowledge systems.
Areas of interest include:
Patent analytics and knowledge representation research
Domain validation and comparative dataset evaluation
Industrial use cases for technology intelligence
Software engineering for research prototypes and analytical systems
Data pipelines, retrieval systems, clustering, graph-based analysis, and AI agent integration
I am particularly interested in collaborating with experienced developers who can help translate research concepts into robust, scalable, and testable systems.
Research Questions
The project is centered on a set of practical research questions:
- Can technical disclosures in patents be decomposed into reusable knowledge units without losing their technical meaning?
- Can these units be grouped into stable technology concepts across different patents and patent families?
- Can relationships among these concepts reveal how technologies evolve, converge, or recombine over time?
- How should such representations be evaluated for consistency, interpretability, and usefulness in real-world technology intelligence tasks?
Where This Could Lead
The research may eventually contribute to new forms of technology intelligence systems in which patents are analyzed not only as documents, but as interconnected representations of technical knowledge.