2025.12 · Pebblous Data Communication Team

Reading time: ~25 min · 한국어

4+
US Registered Patents
24
Korea Applications
3+
PCT Applications
2042
Expiration Year

1. Introduction

1.1. Background and Purpose of the Analysis

As the paradigm of artificial intelligence (AI) technology rapidly shifts from Model-Centric to Data-Centric, securing and validating high-quality data has emerged as a critical factor determining AI performance.

Amid this trend, Pebblous Inc., a Korean deep-tech startup, is presenting solutions that scientifically diagnose and improve data quality through the original concept of 'Data Clinic.'

This report aims to comprehensively survey and deeply analyze Pebblous's intellectual property (IP) portfolio across Korea (KR), the United States (US), Japan (JP), and globally (PCT), thereby elucidating the technological moat they have built and their global market competitiveness.

1.2. Scope and Methodology

This survey was conducted based on data publicly available from patent offices (USPTO, KIPO, JPO, WIPO) as of December 6, 2025.

  • 1. Quantitative Analysis: Application/registration counts by country, patent family size, citation relationship analysis
  • 2. Qualitative Analysis: Interpretation of technology protection scope through claims analysis of key patents
  • 3. Inventor Profiling: Tracking key inventors' research history and analyzing the origins of patent technology
  • 4. Strategic Recommendations: Diagnosis of IP portfolio strengths/weaknesses and global expansion roadmap

2. Executive Summary

2.1. Key Findings

The analysis confirmed that Pebblous has been executing a highly aggressive and sophisticated IP strategy from its earliest days, with the global market, especially the US market, as its top priority target.

US Patent Dominance

Including the latest patent US 12,481,720 registered on November 25, 2025, the company holds a total of 4+ registered patents including US 11,748,447, US 11,868,435, and US 11,967,308. This is the result of a 'Picket Fence' strategy utilizing the 'Continuation Application' system.

Technological Originality

Their patents focus on 'Data Imaging' technology that interprets and visualizes data quality as high-dimensional geometric structures (Manifolds) rather than simple statistics. This creates a technological barrier to entry that is difficult for competitors to replicate.

Global Scalability

Based on priority claims filed with the Korean Intellectual Property Office (KIPO), the company has proceeded with PCT International Application (PCT/KR2023/001001), establishing a bridgehead for entry into individual countries including Japan and Europe.

2.2. Portfolio Summary Table

Category Key Status Notes
US 4+ registered patents (US 12,481,720 et al.) Strong rights secured for data diagnosis and processing. Valid until 2042.
Korea (KR) 24 applications / 4+ registrations Numerous derivative patents filed starting from June 2022 priority claim applications.
Japan (JP) Under examination / publication stage (est.) Expected national phase entry from PCT application.
PCT (Global) 3+ applications and entries Rights reserved (Pending) for entry into major global markets.
Core Technology Data Imaging, Manifold Learning Foundational technology for visualizing data to intuitively diagnose defects.
Inventors Joo-Haeng Lee (CEO), Jeongwon Lee (COO) Convergence of graphics/geometry and brain engineering/bio expertise.

3. In-Depth Inventor Analysis: Origins of the Technology

Patents are the products of inventors' R&D philosophies materialized into legal rights. Pebblous's patent portfolio was born from the complementary expertise of two key figures: CEO Joo-Haeng Lee and COO Jeongwon Lee.

Joo-Haeng Lee

CEO | Ph.D. in Computer Science, POSTECH | Former Principal Researcher, ETRI

Key Research Areas:

  • Computer Graphics
  • Geometric Modeling
  • Computer Vision

Connection to Patents:

Pebblous's 'Data Clinic' interprets data as 'Manifolds' in multidimensional space, which stems from CEO Lee's background in geometric modeling.

Jeongwon Lee

COO | Ph.D. in Bio and Brain Engineering, KAIST | SNU Electrical Engineering/Biomedical Engineering

Key Research Areas:

  • Brain Engineering
  • Biomedical Signal Processing
  • Machine Learning Applications

Connection to Patents:

Experience working with noisy, unstructured data infused practical engineering sensibility into solutions for refining real-world imperfect data.

4. US Patent Portfolio In-Depth Analysis

4.1. Patent Family A: Data Diagnosis and Imaging

The most critical patent group is a series of 'Continuation Applications' related to 'methods and apparatus for diagnosing data attributes.'

US Patent 12,481,720 (Latest Registration)

  • Registration Date: November 25, 2025
  • Application No.: 18/511,617
  • Assignee: PEBBLOUS INC.

Protects technology that projects data into 'first/second embedding spaces' to analyze density and distribution, generating 'data images' to provide visual diagnostic information.

US Patent 11,868,435

  • Registration Date: January 9, 2024 (est.)
  • Application No.: 18/129,711

The parent application of US 12,481,720, including rights to data quality metric computation algorithms and visualization interfaces.

US Patent 11,748,447

  • Registration Date: September 5, 2023
  • Application No.: 17/898,109

The first US application in this family. Filed in the US just 2 months after the Korean priority application, and registered at ultra-fast speed within 1 year.

4.2. Patent Family B: Data Processing and Synthesis

US Patent 11,967,308

Title of Invention: Method and apparatus for processing data for machine learning model

This patent is the legal foundation for Pebblous's 'Precision-Targeted Synthetic Data' generation technology. It identifies holes in data, then uses generative AI to create synthetic data that fills those regions.

4.3. Implications of the US Patent Strategy

1

Rapid Rights Acquisition

Filed US applications immediately after Korean filings, securing priority dates ahead of competitors

2

Chain Grants

Continuation applications from 11,748,447 to 11,868,435 to 12,481,720 block competitor design-around attempts

3

Multi-Layered Claim Scope

Early patents capture broad concepts while follow-up patents cover specific implementations, fundamentally blocking technology appropriation

5. Korea (KR) and Global (JP, PCT) Patent Status

5.1. Korea (KR)

R&D Outpost | Source of Priority Rights

  • Application Scale: Approx. 24 applications / 4 registrations
  • Key Priority Applications:
    • 10-2022-0079508 (2022.06.29)
    • 10-2022-0079509
    • 10-2022-0079510

Concentrated filings within ~7 months of founding - a hallmark of technology-driven startups

5.2. Global Expansion

PCT and Japan (JP) Entry Strategy

  • PCT Application: PCT/KR2023/001001
    • Filing date: January 20, 2023
    • National phase entry: Around July-August 2024
  • Japan (JP): Currently under examination (Pending) estimated

    Expected to be published in H2 2025 or 2026

6. Technology Deep Dive: The Essence of Pebblous Patents

6.1. Traditional DQM vs. Pebblous Data Clinic

Comparison Traditional DQM (1st/2nd Gen) Pebblous Data Clinic (3rd Gen)
Target Data Structured data (text, tables) Unstructured data (images, sensors, 3D)
Diagnostic Method Statistical rule-based Geometric manifold learning
Output Numerical reports (error rate 0.5%, etc.) Data images, visual defect maps
Philosophy "Read data as numbers" "See data as shapes"

6.2. Synthetic Data for Physical AI

In manufacturing sites and autonomous driving environments, 'edge case data that is difficult to generate' is critically important.

Pebblous's patented technology targets the 'holes' in diagnosed manifolds and fills them with synthetic data generated by generative AI. This is not simple augmentation but 'Precision Targeting' technology.

7. Strategic Implications and Future Outlook

7.1. Market Value and IP Competitiveness

Pebblous's IP portfolio corresponds to the 'Pick and Shovel' strategy of the generative AI era.

  • ▪ ISO/IEC 5259 Standard Compliance: The only solution that can quantitatively and visually demonstrate data accuracy, completeness, and other quality metrics
  • ▪ Technology Valuation: The patent registered in November 2025 is valid until 2042, providing a powerful leverage point in M&A negotiations

7.2. Risk Factors and Recommendations

Accelerate Japan/Europe Registration

Patent registration must be expedited in Japan, a manufacturing powerhouse, and Europe, where AI regulations are strict.

Defensive Publication Strategy

A parallel strategy is recommended: patents for core technology, defensive publications for peripheral technology to block competitor entry.

8. Conclusion

Pebblous Inc. has successfully built a global-level IP portfolio based on the deep research capabilities of its two inventors, Joo-Haeng Lee and Jeongwon Lee.

In particular, the chain of patent registrations at the USPTO (US 12,481,720 et al.) clearly demonstrates that they are targeting not just a domestic solution but the global AI data infrastructure market.

Pebblous's technology has preempted a new paradigm through patents: 'Seeing data as shapes (Imaging) and treating it (Clinic).' By around 2026, when patent expansion into the Japan and Europe markets is completed, Pebblous is expected to establish itself as a bona fide 'Global Data-Centric AI' leader.

Frequently Asked Questions (FAQ)

Q1. What technologies does the Pebblous patent portfolio protect?

The Pebblous patent portfolio protects Data Imaging, Manifold Learning, and Precision-Targeted Synthetic Data generation technologies. As of December 2025, the company holds 4+ US registered patents including US 12,481,720, US 11,967,308, US 11,868,435, and US 11,748,447, which form a patent family valid until 2042 through Continuation Applications.

Q2. What is a data quality patent, and what patents does Pebblous hold?

A data quality patent protects technology for diagnosing and improving the quality of AI training datasets. Pebblous's data quality patents include foundational technology for visually diagnosing 'holes' (sparse regions) and density imbalances in training data through Data Imaging, and compensating for them with Precision-Targeted Synthetic Data.

Q3. What is the relationship between the ISO 5259 standard and Pebblous patents?

ISO 5259 is an international standard on data quality for AI training, and Pebblous's patented technology automates verification of the standard's core requirements: 'data completeness' and 'representativeness.' Pebblous's Data Imaging patents can be utilized as dataset quality diagnostic tools for ISO 5259 standard compliance.

Q4. What is the core principle of DataClinic patent technology?

DataClinic patent technology visually diagnoses AI training datasets through Data Imaging, known as 'Data MRI,' and treats discovered defects (holes, outliers, density imbalances) with Precision-Targeted Synthetic Data. The core principle is finding and remedying dataset problems, much like using a medical MRI to locate and treat affected areas.

Q5. Why are Pebblous patents important for Physical AI data quality?

Physical AI (robotics, autonomous driving, manufacturing, etc.) operates in the real physical world, so data quality errors can directly lead to safety incidents. Pebblous patent's 'Precision-Targeted Synthetic Data' technology precisely generates hard-to-obtain edge case data in the 'hole' areas of diagnosed manifolds to reinforce AI model weaknesses.

Q6. What is Data Imaging technology?

Data Imaging is Pebblous's core patented technology that projects high-dimensional data into low-dimensional (2D/3D) embedding spaces for visualization. By mapping semantic similarity between data points to spatial proximity, it enables visual diagnosis of dataset defects (holes, density imbalances, outliers) like an MRI.

References (Works Cited)

  1. Pebblous US Patent Technology and Business Value Analysis Report, accessed December 6, 2025, https://blog.pebblous.ai/project/DataClinic/pbls-patent-us-01.html
  2. accessed December 6, 2025, https://patentsgazette.uspto.gov/week47/OG/html/1540-4/US12481720-20251125.html
  3. US20250238975A1 - Pixel-based image processing method and an, accessed December 6, 2025, https://patents.google.com/patent/US20250238975A1/en
  4. Joo-Haeng Lee - Google Scholar, accessed December 6, 2025, https://scholar.google.com/citations?user=f529BfkAAAAJ&hl=en
  5. Pebblous | AI-Ready Data Solutions: Data Quality, Synthetic Data & Physical AI, accessed December 6, 2025, https://pebblous.ai/en/company
  6. Pebblous | AI-Ready Data Solutions: Data Quality, Synthetic Data & Physical AI, accessed December 6, 2025, https://pebblous.ai/en/traction
  7. PEBBLOUS INC Patent Analysis Report(US Patent), accessed December 6, 2025, https://patent-i.com/report/us_en/applicant/0051467/
  8. S : Press Release Search Results - Newswire, accessed December 6, 2025, https://www.newswire.co.kr/search?sf=1&skey=S&sdate=20201119&edate=20251119&page=95

Report Download

Pebblous Patent Application and Registration Status Survey 2025

Download the full report as PDF for offline viewing.

PDF Download

[Disclaimer] This report was prepared via Gemini based on publicly available materials as of December 6, 2025, and was reviewed by Pebblous CEO Joo-Haeng Lee. Actual legal status may differ depending on unpublished applications or latest examination progress. For legal interpretation of patent scope, please consult a patent attorney or other qualified professional.

Back to Blog