Ryoung Woo

BISTel, Co.

Software Engineer, Data Scientist • Nov. 2013 ~ Present, 5 years 10 months

• Research on new technology of IT and academic paper. Share this with team.
• Gather requirement from external client and internal stakeholder.
• Develop or upgrade a product according to the above technology, paper, requirement.
• Deploy the product to client and Make them solve their problem.

Korea Information Engineering Services, Co.

Internship • Sep. 2013 ~ Oct. 2013, 2 months

• Check quality assurance for product features to be released.
• Refine documentation for a user's point of view.

Daejin University, Ministry of Education, Science and Technology

Completion of course, Hadoop Expert for BigData Storage and Management Based on Cloud Computing • May. 2013 ~ Aug. 2013, 4 months

• Because I had completed this course well, I was able to participate in the internship.

College of Education, Kangwon National University

Bachelor of Science, Mathematics Education • Mar. 2004 ~ Feb. 2012

• Scholarship on 1st year 2nd semester, 3rd year 2nd semester.
• Semi runner up of college basketball competition on 2nd year.

Develop eDataLyzer

Software Engineer, Data Scientist • Nov. 2013 ~ Present, 5 years 10 months, more than 30 M/M

The eDataLyzer is a existing semiconductor analytics product for wafer yield map classification and root cause correlation. The goal of this project is to redevelop the eDataLyzer for big data. So we have led this project in three ways.

• 1st, reconstruct the architecture to micro services from monolithic one.
• 2nd, reorganize to a role based teams from a unified team.(Client, Server, Algorithm, Research, Technical Sales/Support.)
• 3rd, redevelop by Java and C# not only C#.

I belong to Algorithm team, mainly focusing on parallelizing algorithm by new big data technologies. This project had three subproject. In briefly, I have done three ways of parallelizing algorithms.

• 1st, I had redeveloped the algorithms by Java, PostgreSQL, Spring for small data clients.
• 2nd, I had redeveloped the algorithms by non Hadoop based technologies.(GreenPlumDataBase PL/Java, Oracle-R)
• 3rd, I had redeveloped the algorithms by Hadoop based technologies.(Hadoop, BDA, Hawq, HBase, Spark, Eco system)

The clients of this project are various as follows. And the project types are PoC, Pilot and Production.

• Korea - Samsung Electronics, SKHynix, SKSiltron
• Japan, Taiwan, China - Toshiba, Sharp, TSMC, BOE

Research to Apply Reinforcement Learning on Semiconductor

Researcher, Software Engineer • Dec. 2018 ~ Present, 9 months, 6 M/M

The goal of this project is to apply reinforcement learning on semiconductor and share the experience with team. I pick up 8-Puzzle as a environment for reinforcement learning. Here's why I pick up it in details.

• In order to collaborate with teammates, I need to find the generalized environment easy to apply Graph Theory. So I pick up operation management of production on semiconductor.
• In order to find suitable environment not complex one, So I pick up 8-Puzzle. Focus on shortest-path not yield, productivity, stability, automation rate, etc

The recent research situation is as follows.

• 1st, solve 8-Puzzle by Dynaminc Programming
• 2nd, fail to solve 8-Puzzle by QLearning, Deep SARSA, Polish Gradient

The rest of the research is to find out why and how to overcome it.

Develop Matrix Profile on Transfer System

Software Engineer • Jul. 2017 ~ Dec. 2017, 6 months, 2 M/M

The goal of this project is to predict a shutdown of motor based on time series sensed data. Client's product-lines(blue-collars) found out a shutdown once a year and hate this problem. But both client's office-lines(white-collars) and our previous algorithm didn't predict it. Because the algorithm focus on a vibration analysis on rotationary machine. So we created this project as a subproject of the previous one and led this project as below.

• 1st, we found out the matrix profile which is a suitable algorithm for time series predict.
• 2nd, I implemented the algorithm in python and deploy it to client. And solve the problem successfully.
• 3rd, I implemented it in java and integrate with UI. Teached them how to solve their problem through our product.

The client of this project is Hyundai Motor. And the project type is PoC.

Develop Predictive Maintenance on Semiconductor

Software Engineer • Nov. 2016 ~ Mar. 2017, 5 months, 4 M/M

The goal of this project is to provide predictive maintenance on semiconductor's etching tools. In order to change this client's empirical maintenance by condition and time, we led this project as below.

• 1st, redefine input data through Self Organizing Map. And define health scores from distance of each vectors of input data and observation vector.
• 2nd, apply Double Exponential Weighted Moving Average to the health scores. And get Remaining Useful Life for each vectors of input data.

But we got a feedback that our product was too late for client. So we additionally led this project as follows.

• 1st, detect the bottle neck of our product as the DEWMA not the SOM.
• 2nd, apply Spark and HDFS on 2nd DEWMA algorithm. And find out a tuning point.

The client of this project is SKHynix. And the project type is pilot.

QA & Documentation on new product

Internship • Aug. 2013 ~ Oct. 2013, 3 months, 5 M/M

During the internship, I did the following activities.

• 1st, quality assurance for each feature on new product.
• 2nd, modify the previous document by the user's perspective.

Programming - Java, SQL, Python

I have used Java, SQL for 6 years on production. I have used Python for 2 years on papers research.

Backend - Spring Frameworks

I have used this for a year on production. Using this, we can re-construct the previous monolithic application into micro services architect. Using this, we can re-build the previous configure hell party into AOP based application.

Database - Oracle, PostgreSQL, HBase

I have used Oracle for a year on production. Because it's used by most semiconductor clients. I have use PostgreSQL for 6 years on production. Because it's a alternative of Oracle. It's easy to use and especially good for performance over price. I have used HBase for a year on production. Because it's a alternative of Oracle. It's hard to use but better than PostgreSQL especially on performance.

Bigdata - GPDB, Hadoop, Spark

We realized that it takes a long time to extend my previous product to big data with Java alone. In order to reduce the time, we find out the following Bigdata platform which enable us to focus on business logic of the product. 1st, GPDB PL/Java is a static platform which enable us to handle data by SQL, compute data by Java. 2nd, Hadoop Spark is a dynamic platform which enable us to handle and compute data by Java. Using Hadoop with other platform not Spark, we couldn't overcome PoC. I have used GPDB for 6 years. I have used Hadoop, Spark for 4 years.

ML, AI - Nvidia CUDA, Keras, Tensorflow

I have used above skills for Time series data mining and Reinforcement Learning project for a year on papers research.

DevOps - Linux, Docker, On premise Cloud(KVM), Public Cloud(Azure, AWS)

Use these for utility tool on projects.

Research

Read and implement the latest papers by python.

Leadership

Lead successfully both projects for a year.

Communications

Communicate smoothly with various stackholders (aggressive client, academic advisor for govermenti, etc).

Language

Native in Korean, Limited working proficiency in English.