Yoshio Takaeda | Portfolio

Forward Deploy Engineer / Data Architect / Researcher / Lecturer

Yoshio Takaeda

Data Engineer / Researcher / Lecturer
Working with manufacturing data to develop predictive maintenance and stability control systems.

Let AI organize the chaos; humans courageously add dimensions.


Overview

With over 30 years of experience in manufacturing process design, data engineering, and analytics infrastructure development. From chemical plant design to genome analysis engines, polymer material databases, and current high-dimensional time series analysis—consistently pursuing data-driven problem solving.

Currently serving as a lecturer at Sophia University Graduate School, teaching “Practical High-Dimensional Data Analysis” and bridging theory with implementation.


Experience

Education

Kyoto University Graduate School of Engineering / Ph.D. (Engineering)
Polymer Chemistry, Polymer Molecular Theory Laboratory (1995)

Professional Experience

Sophia University Graduate School, Applied Data Science Program / Lecturer
September 2023 - Present

toor Inc. / Founder & Chief Research Officer
January 2012 - Present

Public Tech Company / Chief Technology Officer
August 2006 - January 2012

Mitsubishi Research Institute / Senior Research Scientist
July 1998 - July 2006

TonenGeneral Sekiyu Chemical / Research Engineer
April 1995 - June 1998


Technical Expertise

Core Skills

30+ Years of Data Engineering in Manufacturing
From chemical plant design to genome analysis, polymer material databases, and manufacturing equipment predictive maintenance—consistently pursuing data-driven solutions

High-Dimensional Time Series Analysis
Handling hundreds to thousands of manufacturing process parameters as multidimensional vectors, enabling seamless (continuous) visualization of complex system state transitions across optimal time scales for dynamic characterization and predictive maintenance

Acoustic Signal-Based Predictive Maintenance
High-quality audio data acquisition using DSP processing with Raspberry Pi + ADC + PreAmp + piezoelectric sensors to detect anomalies in rotating machinery and other manufacturing equipment

Data Infrastructure Design
SQLite3 for rapid POC validation, PostgreSQL/TimescaleDB for production-scale time series data processing. Emphasis on appropriate technology selection

Algorithm Development
Homology search (genomics), material property prediction (polymers), anomaly detection (manufacturing processes and acoustic signals)

Technology Stack


Research Projects

🧭 toorPIA (High-Dimensional Time Series Analysis Engine)

Research project for real-time analysis of manufacturing equipment parameters, automatic anomaly detection, and root cause identification.

Research Focus

Technology Stack
Python API (signal processing & analysis), BFF (Node.js + React/Next.js), PostgreSQL + TimescaleDB, Docker/Kubernetes

→ GitHub: toorpia/toorpia

🔊 Acoustic-Based Predictive Maintenance System

Research project for acquiring audio data from rotating machinery and other manufacturing equipment to detect anomalies.

Technical Architecture

Proof-of-Concept Results

🪶 Local LLM Analysis

Experimental research project analyzing internal representations of local large language models to visualize proposition distances and bias structures.

→ GitHub: toorpia-labs/local-llm-analysis


Research Chronology

2012-Present: Manufacturing Predictive Maintenance (toor)

Consolidating 30 years of experience to develop technology that keeps manufacturing sites running. Dual approach using both process parameters and acoustic signals.

2006-2012: IP Management Systems (Venture Company)

Solving the complexity of intellectual property management through data infrastructure. Integration of technology and business.

1998-2006: Genomics & Materials DB (Mitsubishi Research Institute)

Large-scale data processing in the International Human Genome Project, structured data design for polymer materials database.

1995-1998: Chemical Plants (TonenGeneral)

Process design and data analysis for PP/PE polymer manufacturing plants. The foundation of manufacturing data engineering.


Patents & Publications

Patents (United States)

Multidimensional Correlated Data Extracting Device and Method
US Patent 10,353,892 (2019)
Method for extracting correlated subsets and identifying featured elements in multidimensional space

State Determining Device and Method
US Patent 10,621,028 (2020)
Time-series combination of device data and similarity-based state analysis

Data Analysis Apparatus and Method
US Patent Application 20160109355
Segment analysis using similarity indices based on frequency spectrum

Document Retrieving Apparatus and Method
US Patent 8,818,979 (2014)
Interactive document retrieval and query vector combination on 2D maps

Analyzer, Analysis System, and Analysis Method
US Patent Application 20210232567
Method for plotting new data on reference maps

Information Display Method and Device
US Patent Application 20170213249
Advertisement display clarifying relevance in search services

Academic Publications

Genome Analysis

Polymer Solution Light Scattering & Nuclear Magnetic Relaxation Studies (1993-1997)


Research Philosophy

Field-Oriented Approach
Pursuing structures that “continue to operate” rather than just theory. Always staying grounded in the field—from chemical plants to genome analysis to manufacturing equipment.

Appropriate Technology Selection
Avoiding over-engineering. Start small and iterate quickly.

Critical Thinking
Incorporating opposing viewpoints to surface risks early. An attitude learned from 30 years of failures and successes.

Learning Cycle
Taking technology back to the field and learning from the field again. Mutually reinforcing education, practice, and research—emphasizing the trinity of these three elements.

Public Individual Statement I believe that individual engineers should not be consumed by corporate structures, but instead deliver their value directly to the fields that truly need them.


Contact

GitHub: github.com/takaeda
Email: takaeda@gmail.com
Location: Japan

This site showcases personal research and educational activities and is non-commercial.


© 2025 Yoshio Takaeda | Personal Research Portfolio