Project Portfolio: Jeayoung Jeon

MLOps/DevOps and AI Engineer (Updated at 2024-12-31)

My name is Jeayoung Jeon [전제영], and I’m an MLOps engineer in Seoul, South Korea. I also specialize in:

🐳 Developing MLOps (APIs, Pipelines) and AI/LLM Platforms in cloud-native environments.
🧑‍🔧 Building Hybrid Kubernetes Clusters for High Availability and GPU Cost Reduction.
🧑‍🎓 Contributing decisions for MLOps/DevOps using backgrounds in ML, Computer Vision, Automotive.

I’m always open to new challenges and opportunities for various fields including ICT, AI, and Automotives. Please feel free to contact me. If you’re looking for my brief resume and works, please see my resume and portfolio.

PDF

Email: jyjeon@outlook.com
Website: https://jyje.live
LinkedIn: LinkedIn: jyje

GitHub: Github
StackShare: StackShare

Work

Mar 2024 – Nov 2024 (9 Months)

🧑‍🔬 Intermediate Software Engineer [책임연구원] at MAXST

Roles: Lead MLOps/DevOps Engineer at Technology Division, MAXST

MLOps Developing ML APIs, data pipelines, and AI Platforms for research team using open sources.
LLMOps Building chatbots using self-hosted RAG+LLM systems for internal documents.
SRE Site reliability engineering for web services. Service reliability engineering for ML workloads.

Jan 2021 – Feb 2024 (3 Years and 2 Months)

🧑‍💻 Software Engineer [선임연구원] at MAXST

Roles: Associate Researcher and DevOps Engineer at Technology Division, MAXST

Algorithm Research Reviewing computer vision algorithms in state-of-art papers and implementing prototypes
Hybrid Clusters Building hybrid clusters with AWS EKS and on-premise Kubernetes for digital twin project
DevOps Building on-premise clusters and data pipelines for company's inbound/outbound projects

Mar 2012 – Aug 2020 (8 Years and 6 Months)

🧑‍🎓 Graduate Student Researcher in Computer Vision at POSTECH

Roles: Ph.D Integrated Student at Department of Electrical Engineering, POSTECH

Computer Vision Research on hyperparameters for accurate and efficient computer vision algorithms
Automotives Principal computer vision technologies for autonomous driving including ADAS and SLAM; Participated in the development of the Korean government's ADAS research projects
FPGA Efficiently implemented computer vision and machine learning algorithms with real-time parallel matrix processing; SoC-type GPU/NPU accelerator

Projects

Jan 2024 – Oct 2024 (10 Months)

🧑‍💻 Widearth: Digital Twin & AR Content Platform at Widearth, MAXST

Roles: Lead ML/Infra Roles ~ MLOps/DevOps + ML Backend + SRE [contrib 75%]

DevOps & SRE IaC, GitOps, CI/CD Pipelines, Monitoring, Logging, Notifications, Multi-Deployment, Emergency Response
Hybrid Clusters Public Cloud + On-Premise Kubernetes, API Gateway Pattern, Dynamic VMs, GPU Cost Optimization
ML Workloads ML APIs, ML Pipelines, Data Lakes, Dockerizing, Model CI/CD

Results: Service Launch ~ Small Team, Full Features, More Availability, Less Cost

Launch Launched/Operated a platform as 1 infra engineer with 15 people, 8 developers in 10 months.
Low Cost Reduced cloud costs by 15M KRW (70%) by using hybrid clusters for 300+ maps.
Robust Infra Achieved 96% availability/year and 14d downtime using hybrid clusters and damage control.

Skills: Skill Stack for Project Widearth

AWS EKS
Kubespray
Python/FastAPI
Argo Workflows
Argo CD
Bitbucket Pipelines
Karpenter

Jan 2024 – Jun 2024 (6 Months)

🧑‍🔬 MLOps: On-Premise MLOps with Open Source Projects at MAXST

Roles: Lead MLOps Engineer ~ Planning + VoC + PoC + ML Workloads/Infra + Operation [contrib 90%]

Kubeflow Integrated Argo Workflows; AutoML, Distributed Training, Model Registry; 16 GPUs Acceleration
JupyterHub Integrated JupyterHub with IDE; Remote GPU Notebook, 4 GPUs Acceleration
VectorDB Milvus, ChromaDB, RAG+LLM Chatbot
ML Infra Setup CI/CD, NAS, Data Lake, Image Registry for ML Workloads

Results: Improved research environment and resource management ~ Increased availability and capacity by merging resources and automating management.

Improved Env. Consolidated servers managed by researchers into k8s to stable infra capacity and stability; Decision-making using PoC.
AI Platform Expanded from VoC of 2 researchers, gradually increased users to 10. Resolved technical debt through continuous MLOps upgrades.
GPU Utilization GPU usage increased by 3 times and successfully commercialized as a result of performing over 800 AutoML experiments

Skills: Skill Stack for On-Premise MLOps

Kubeflow
Katib
Training Operator
Model Registry
JupyterHub
Argo Workflows
Milvus
ChromaDB
Ollama
Open WebUI
Grafana Stack
TensorBoard

Dec 2022 – Dec 2023 (13 Months)

🧑‍🔧 DevOps: Hybrid Clusters for Internal/External Projects at MAXST

Roles: DevOps Engineer ~ Hybrid Clusters + CI/CD + Chatbot + Data Pipelines [contrib 50%]

Hybrid Clusters Public Cloud, On-Premise Kubernetes, Multi-Cluster, API Gateway, IaC, GPU Operator
CI/CD Public CI Platform, On-Premise Custom CI, GitOps CD, ChatOps for Results/Issues
Pipelines Data Pipelines for ML Research, Production Pipelines for ML Inference

Results: Hybrid Cluster Initiation ~ Increased On-Premise Resource Utilization + Reduced Public Resource Costs + DevOps Culture Propagation

Cost Reduction Maintained public availability while reducing costs by 50% compared to pure cloud infrastructure using on-premises cost-effectiveness.
Resource Utilization Utilized 90% of idle on-premises resources, provided multi-cluster for prototyping in other departments
DevOps Culture Introduced cloud-native development environment. Propagated DevOps culture including app modernization and CI/CD. Decision support through monitoring.

Skills: Skill Stack for DevOps and Hybrid Cluster

Kubernetes
AWS EKS
IaC
Ansible
Terraform
CI/CD
Bitbucket Pipeline Runners
Argo CD
Argo Workflows
Python/FastAPI
Python/Bolt (Slack)

Jan 2021 – Dec 2022 (2 years)

📸 Computer Vision Engineer at MAXST

Roles: Associate Researcher ~ Algorithm research for digital twin systems and prototyping [contrib 50%]

Digital Twins Digital twin system implementation using algorithms for converting perspective and 360 images to 3D space.
AR/XR Camera calibration and AR/XR prototype development for various smart glasses
Automation Development of automated pipelines for data acquisition and analysis
Military Service Engaged in position related to graduate school majors and performed alternative military service.

Results: Development of computer vision algorithms and construction of digital twin systems

Digital Twins Research and development of Visual-SLAM and ICP algorithms for digital twin systems
Automation Development of automated pipelines for data acquisition and analysis

Skills: Skill Stack for computer vision research

Computer Vision
Visual-SLAM
SfM
ICP
Python
OpenCV
.NET/C#
Unity

Jan 2012 – Aug 2020 (8 Years)

🧑‍🎓 Computer Vision and ADAS Researcher (Integrated Program) at POSTECH

Roles: Graduate Student Researcher ~ Computer Vision and ADAS Research [full-time]

2018-2020 Computing and Control Engineering Lab. (Prof. SH, Han)
Digital Twins and Simultaneous Localization and Mapping (SLAM) Research
- Visual-SLAM Research using Multiple Cameras for Autonomous Driving
- Prototyping of Digital Twins for ADAS and SLAM
- Virtual Visual-SLAM for Real-World Environments
2012-2018 Advanced Signal Processing Lab. (Prod. H, Jeong)
Advanced Driver Assistance Systems (ADAS) and Edge Computer Vision Research
- High-Performance, Efficient FPGA Implementation of ADAS
- High-Speed Algorithm Development for Traffic Signs and Road Terrain Detection
- Research on Stereo Vision Algorithm for 3D Depth Estimation
- Stereo Vision-based Online Calibration for Vehicle Cameras
- Optimization Algorithm Research for Computer Vision using Cost Aggregation Table

Results: Proejcts and Research Papers ~ Studying on Automotive Simulations in Virtual Environments and ADAS On-Edge.

Digital Twins Virtual Visual-SLAM for Real-World Environments
Edge ADAS Research of ADAS including Traffic Sign Detection & Lane Terrain Detection with FPGA

Skills: Skill Stack for Computer Vision and ADAS Research

Computer Vision
Digital Signal Processing
Automotives
Autonomous Driving
Advanced Driver Assistance Systems (ADAS)
Finite Programmable Gate Array (FPGA)
Traffic Sign Detection
Lane Terrain Detection
MATLAB/Simulink
C/C++

Skills

Here are my skills and highlighted items are industry-ready.

MLOps & LLMOps

Ollama
OpenAI API
RAG
AutoRAG
Kubeflow
AutoML Katib
Training Operator
JupyterHub
Data Pipelines

DevOps & SRE

Kubernetes
On-Premise
AWS EKS
GCP GKE
Hybrid Clusters
ARM64
IaC
Kubespray
Terraform
Ansible
Istio
Grafana Stack
Karpenter

CI/CD/CT/CT

Argo Projects
Bitbucket Pipelines
GitHub Actions
Self-Hosted Runner
Kaniko
Buildah
Locust
Litmus

ML Backend

Python/FastAPI
Ollama
Milvus
PostgreSQL
Redis

Computer Vision

Automotives
SLAM
PyTorch
OpenCV
FPGA

UI/UX

Slackbot
Python/FastUI
Open WebUI
Chainlit
.NET/MAUI
.NET/WPF
Unity

FinOps & BizOps

Kubecost
Continuous BI

Programming languages

Python
.NET/C#
C/C++
MATLAB

Education

Mar 2012 – Aug 2020

🎓 Master's Degree (Integrated Program) in Department of Electrical Engineering, Signal Processing & Computer Vision from Pohang University of Science and Technology (POSTECH) with GPA of 3.2/4.3

- Thesis: Virtual Visual-SLAM for Real-World Environments, 2020

Mar 2008 – Feb 2012

🎓 Bachelor's Degree in School of Electronic Engineering, Electronic Communication from Kumoh National Institute of Technology (kit) with GPA of 4.3/4.5

- Thesis: A Study on a Visible Light Communication using LED in Under-water Environment, 2011

Awards

May 2014

🏅 Altera Design Contest 2014, Excellence Prize from Intel-Altera Korea

[System] FPGA, Vision-Based Driver Support Navigation System

May 2014

🏅 Best Poster Session in Workshop from KYUTECH-POSTECH Joint Workshop

[Poster] Iterative Polygon Detection using Harris Corner Space Method for Finding Traffic Signs

May 2013

🥈 Altera Design Contest 2013, 2nd Prize from Intel-Altera Korea

[System] FPGA, Vision-Based Traffic Sign Recognition System

Feb 2012

🏅 Highest Honors in Undergraduate School from Kumoh National Institute of Technology

[Summa Cum Laude] Highest Honors in Undergraduate Electronic Engineering School

Jan 2012

🏅 NAVER Power KiN 2011 from NAVER

[Activity] Knowledge Export in `Electronics Engineering, Mathematics and Programming fields`. Active 2009-2011, Selected as a MVP in 2012 / Total number of answers 723, Selection ratio 98.1%