DoangJoo "Alan" Synn

I am a PhD Student in the fields of Computer Science at Georgia Institute of Technology. My research interests span Distributed systems, Cloud Computing, and System for AI. Before I get into PhD program, I worked as a software engineer for 8+ years in various layers of distributed system, machine learning system and cloud computing.

During the period, I was a Research Scientist at Protopia AI inc. which is a AI company mainly tackles for privacy in Austin, Texas. In Protopia AI, I develop and maintain internal frameworks and libraries to support privacy-preserving machine learning with distributed ML frameworks and cloud computing. And I was a AWS Educate Cloud Ambassador of Amazon Web Service. I have profound experience developing distributed frameworks on cloud, models using PyTorch, and deploying models using ONNX.

I received M.Eng and B.Eng in Electrical and Computer Engineering from Korea University advised by Prof. Jong-Kook Kim. My research at Korea University focused on Machine Learning Systems and Distributed Systems.

General Information

Full Name	DoangJoo (Alan) Synn
Contact	alansynn [at] gatech [dot] edu
Office	Klaus Advanced Computing Building 3110
Languages	Korean (native), English (fluent), Japanese (beginner)

Research Interests

My research interest focuses on developing distributed system and Machine learning system, especially for efficiency and scalability.
I have two directions for my research; (1) build scalable system to serve the cutting-edge workload and (2) Build systems for machine learning to accelerate

Education

2022 -
Georgia Institute of Technology – PhD in Computer Science
- Advised by Prof. Alexey Tumanov
2020 - 2022
Korea University – M.Eng in Electrical and Computer Engineering
- Advised by Prof. Jong-Kook Kim
- GPA: 4.25 (4.5 scale)
- Thesis: On Efficient Data Delivery for Machine Learning Systems (ongoing)
2012 - 2020
Korea University – B.Eng. in Electrical Engineering
- GPA:3.56 (scale 4.5)
- on leave from 2016 Spring to 2019 Spring to serve alternative military service

Work Experience

2021 - 2022
Research Scientist @Protopia AI inc.
- Austin, Texas, USA
- CEO: Dr. Eiman Ebrahimi
- CTO: Prof. Hadi Esmaeilzadeh
2019 - 2022
Cloud Architect
- Design cloud native softwares and consulted 4 different companies to manage the DevOps infrastructure
- Field : Cloud Computing, Cloud Applications, Data Pipeline
2020 - 2022
Amazon Web Services(AWS) Educate Cloud Ambassador
- The AWS Educate Cloud Ambassador Program is for top-tier cloud educators who serve as ambassadors and evangelists for AWS Educate.
2019
Server Engineer @Flysher inc.
- Global #10 social game
2017-2019
System Engineer @AKA Intelligence inc.
- Develop and Produce Embedded board & S/W for AKA INTELLIGENCE's social robot, Musio.

Honors and Awards

Fellowships
- Brain Korea 21(BK21) Research Fellowship (2020-2021)
Awards
- Prize Winner Creative Challenge Program @Korea University Center of Teaching and Learning (2016)
- Samsung Software Friendship @Samsung SDS (2016)
- Prize Winner Creative Challenge Program @Korea University Center for Teaching and Learning (2015)
- EE Academic Excellence Honors Program (2012, 2019)

Teaching Experience

Spring 2020
Teaching Assistant for Digital System
- Undergraduate level introduction to digital system class at Korea University
- Taught by Prof. Jong-Kook Kim
2012 - 2014
Teaching experience at KAIST engineering school
- Develop the software education curriculum for the student bootcamp conducted by Ministry of Science ICT and Future Planning in Korea.
- Taught software engineering and programming for middle school and high school students.

Technical Projects

Privacy-Preserving Deep Learning @ Protopia AI
- Protopia generates curated noise around the source signal, safeguarding data from unauthorized inference.
  Illustration of our technology on Mask Detection task on Baby Faces
  Demo video for our technology on Face Detection task
  NetApp Post: Responsible AI and ethics by design: Is your privacy governance-ready for AI at scale?
  
  Tasks at Protopia Ai inc. :
  Development of Protopia AI’s noise training technology for inference privacy.
  
  Development of Internal framework and system for noise training technology for privacy.
  
  Research on Face Recognition System with Privacy Preserving with noise training
  
  Development of Scalable training and Workflow management system for noise training
  (2021.1 – 2022.8) NeurIPS-D’21
High scalable node.js game server framework @Flysher inc.
- Game server framework for Typescript developers which provides the basic development framework and many related components, including libraries and tools
- The former framework, which is pomelo has been deprecated. Therefore, we need to make a new framework based on the former architecture in order to achieve fast and adoptable framework.
Musio @AKA Intelligence inc.
- Musio is a social robot designed by AKA Intelligence, powered by MUSE, a deep learning based AI quantum engine for communication developed by AKA. Using deep learning algorithms, Musio converses with people, recognises objects and also understands people's facial expressions.
- For two years, I was an engineer at AKA Intelligence inc., where I worked on running various applications on embedded devices. I tackled the problem from various levels of the system stack.

In training a deep learning model, it is crucial to tune various hyperparameters and gain speed and accuracy. Although the hyperparameters that mathematically affect the convergence significantly affect the training speed, the system parameters that affect the host-to-device data set transmission time also occupy a specific part in the overall time acceleration. Therefore, it is important to properly tune and select parameters that can affect the data loader as a system parameter in overall time acceleration. We propose an automated framework called Num Worker Tuner (NWT) to address this problem. This method finds the appropriate number of multi-processing subprocesses through the search space and accelerates the learning through the number of subprocesses. Furthermore, this method allows memory efficiency and speed-up by tuning the system-dependent parameter, the number of multi-process spawns.

Many natural language processing (NLP) models utilize pre-trained word embeddings to leverage latent information. One of the most successful word embedding model is the Skip-gram (SG). In this paper, we propose a Skipgram drop (SG-Drop) model, which is a variation of the SG model. The SG-Drop model is designed to reduce training time efficiently. Furthermore, the SG-Drop allows controlling training time with its hyperparameter. It could train word embedding faster than reducing training epochs while better preserving the quality.

DoangJoo "Alan" Synn

General Information

Research Interests

Education

Georgia Institute of Technology – PhD in Computer Science

Korea University – M.Eng in Electrical and Computer Engineering

Korea University – B.Eng. in Electrical Engineering

Work Experience

Research Scientist @Protopia AI inc.

Cloud Architect

Amazon Web Services(AWS) Educate Cloud Ambassador

Server Engineer @Flysher inc.

System Engineer @AKA Intelligence inc.

Honors and Awards

Teaching Experience

Teaching Assistant for Digital System

Teaching experience at KAIST engineering school

Technical Projects

Privacy-Preserving Deep Learning @ Protopia AI

High scalable node.js game server framework @Flysher inc.

Musio @AKA Intelligence inc.

Publications