Backend Engineer @ Samsung Electronics
- I build and operate the backend of a backup and restore service for tens of millions of users at Samsung Electronics. (Java, Spring Boot, AWS, Kafka)
- I use operational logs to measure user experience. I find failure causes and improve service behavior and operations.
- I automate repetitive analysis with AI agents. I use tests and code reviews to check existing behavior after changes. I hold the in-house Best Reviewer certification.
# experience
Samsung Electronics
Samsung Electronics
Backend Engineer
Jan 2023 – Present · Suwon
I develop and operate the backend of a cloud backup and restore service used by tens of millions of people. My work covers service features, operational log analysis, quality metrics and tools for everyday workflows.
Responsibilities
- I build APIs and business logic with Java and Spring Boot, and operate an AWS-based microservice environment connecting databases, storage and messaging.
- I process large volumes of operational logs to calculate user-perceived quality metrics, find failure causes and improve backup features and operations.
- I build VOC lookup tools, shared log query tools and AI agents to reduce repetitive analysis and response work. I check behavior after changes with tests and code reviews.
Shinyoung Securities
Software Engineering Intern · IT Planning
Sep 2021 – Dec 2021 · Seoul
During the mobile trading system renewal, I first added a Node.js test server and checks. I migrated screens and modules in stages, comparing results with the previous version at each step. I documented test criteria and differences for the team.
MINDsLab (now maum.ai)
Software Engineering Intern · Biz AI
Jul 2020 – Dec 2020 · Pangyo
Backend maintenance for an AI identification/tracking system — test server monitoring and incident response
# work-projects
Completed while at Samsung Electronics · Jan 2023 – Present
User-Perceived Backup and Restore Quality
I grouped API calls into complete operations to measure whether users actually finished their backups and restores.
- Defined success by the final outcome of each user operation. I checked the results against operational logs.
- Processed logs with Spark and calculated daily metrics with Airflow. Results were available in Grafana. Failed tasks could be rerun without restarting the entire pipeline.
- Used metrics by failure stage, region, and API to find causes and improve service behavior and operations. Daily user-perceived backup success rose from about 70% to about 95%.
Daily user-perceived backup success rate ~70% → ~95%
SLI/SLO Anomaly Analysis and Reporting Automation
When a metric fell below its threshold, engineers gathered logs from multiple servers and compared them with server code to investigate the cause. I automated this analysis, which took up to four hours, by dividing it among three agents.
- The log extraction agent collects logs matching the alert's service, region, API, and time window. It groups logs from the same transaction and selects those needed for analysis.
- The cause analysis agent uses RAG to find similar past incidents. It compares them with supporting logs and server code to narrow down possible causes.
- The reporting agent posts possible causes and evidence in the original alert thread. Engineers can request analysis and review the evidence where they received the alert.
- I reviewed early results and updated agent instructions after incorrect conclusions. I saved reviewed error patterns for future analysis. Direct measurements of the same task showed a reduction to about ten minutes.
Analysis and reporting time: up to 4 hours → ~10 minutes
Building a Web Tool for Backup VOC Analysis
Client developers had to ask server developers for logs to investigate customer issues. Repeated requests delayed resolution, so I built a web tool that lets them look up the information directly.
- Selected five lookup functions that client developers commonly used. I handled the design, frontend, and backend development.
- Users enter user information, a time period, and a lookup function. The backend pipeline calls internal APIs or runs Athena queries through Lambda and displays the results.
- Added filters and Markdown export to support analysis and response. Client developers can now retrieve the information and logs they need without going through server developers.
VOC resolution: at least a day → within a day
Mobile Backup E2E Test Automation and Daily Quality Reporting
I automated real-device E2E verification tests that were previously run and compiled manually every day, and built daily reports and a dashboard to track failure causes and quality trends by device.
- Analyzed client code to implement core verification scenarios as automated real-device tests. Built a daily pipeline that runs tests on schedule, collects log and video evidence, masks personal data, classifies failure types, and generates reports.
- Visualized accumulated test results as a device-and-test heatmap web dashboard to identify and fix flaky tests on specific models.
- Built a bot workflow where teams trigger test runs via messenger mentions and check progress and results in threads, now used across the backup organization and partner teams.
Automated daily real-device testing & reporting (privacy check: 18 min → 2 min)
AgentHub: AI Agent Session Manager
I built a tool to monitor agents across multiple terminals, spot input and approval waits, and resume work from one screen.
- Grouped tmux sessions by working directory. I highlighted states that needed attention or intervention.
- Connected users directly to each session terminal to enter instructions and continue their work.
- Handled planning, Electron and React app development, and internal Tool Store deployment independently. Other developers use the tool in their work.
Ranked #3 in recommendations among 15 tools
A Shared Log Query Tool with Jupyter and PySpark
When support for the existing log analysis tool ended, teammates had to write their own queries in Jupyter. I built a shared interface that retrieves logs from a few input conditions.
- Teammates enter a module, region, time period, and keyword to query S3 logs with PySpark. This reduces the work of finding log paths and repeatedly writing query code.
- Flattened nested fields into columns and provided keyword filtering. Teammates can use the results directly in further analysis.
- Demonstrated the tool in a team knowledge sharing session and shared the documentation and notebook. I incorporated feedback from actual use through version three.
Best Reviewer Certification and Code Quality Improvements
I earned the in-house Best Reviewer certification through Clean Code, Refactoring, and TDD training, practical application, and an interview.
- Checked existing behavior with tests, then separated mixed responsibilities while gradually refactoring legacy code.
- Created a code review template covering reasons for changes, affected behavior, and testing methods. I shared recurring review criteria with the team.
- Created AI code review guidelines and agent skills used in development and reviews, with tests to check how new features and changes affect existing behavior.
Automating Vulnerable Package Analysis and Update PRs
I built Security Bot to keep up with Dependabot security issues across legacy repositories. It automates issue collection, version analysis, and update PR preparation.
- GitHub Actions periodically checks repositories and groups security issues by package.
- The LLM analyzes CVE details, current versions, and package configurations to suggest a target version.
- When the target version is clear, a script updates package configuration and creates a PR with a review request. Cases requiring a decision include candidate versions and supporting reasons.
- Maintainers review changes and test results before merging. I also built a web dashboard showing issues and PR status across repositories. The bot and dashboard remain in regular use.
# community-activities
Learning, building and sharing with others
-
SIPE
5th cohort · Design system (side) · Linktrip developer
In the first mission of the 5th cohort, I implemented semantic tokens, accessibility hooks, and responsive Flex components for the in-house design system (@sipe-team/side) and migrated layouts on the official website. In the second mission, I developed place candidate filtering and shared map views for Linktrip, a service that extracts places from group chat links.
-
SSAFY
8th cohort · Study group lead · Mentor
I studied Java, Spring and algorithms in the 8th cohort's Java track and led an algorithm study group. I built, deployed and operated an attendance excuse generator, and now share learning and development experience as a mentor.
-
GDSC CAU
Web development · MOA developer
I studied web development weekly and shared development experience. I planned and built the delivery mate matching service MOA on my own, including the frontend, backend, database and deployment, and incorporated feedback from club members.
-
Pirogramming
14th cohort · DOWITH team lead · Session speaker
I studied Python and Django web development in the 14th cohort and led planning, database modeling, development, and AWS EC2 deployment as team lead for DOWITH, a habit-building challenge service. After graduating, I led Django environment sessions for the 15th and 16th cohorts and AWS EC2 deployment sessions for the 17th and 20th cohorts.
-
UMC
2nd cohort Node.js track · Team project backend developer
I studied layered REST API design and server development with Express and MySQL in the 2nd cohort's Node.js track at Chung-Ang University. With my track teammates, I planned a character personality test web service timed with a drama release and built and deployed the result-matching API using Node.js and AWS RDS.
-
SOPT
26th cohort Planning track · Booster planner & PM · 16th APPJAM Grand Prize
I led planning and project management for Booster, a print pre-order app, at the 16th APPJAM. I defined user flows, priorities and schedules, coordinated developers and designers, and won the grand prize.
# side-projects
Projects I planned and built
SSAFY Attendance Excuse Generator
I built a service that generates attendance excuse PDFs from form inputs, simplifying a process that required printing, handwriting, and scanning. I implemented it with Vue and deployed it on AWS EC2. It ran for about three years and was used over 10,000 times.
A web service where people build daily routines together instead of giving up on plans alone. As team lead of a five-person team, I led planning, database modeling, friend and challenge feature development, and AWS EC2 deployment.
A service connecting dorm residents ordering from the same restaurant to share delivery costs and meet minimum orders. I handled planning, the JavaScript frontend, Django and MySQL backend, and AWS EC2 deployment independently. I improved the service based on feedback from actual users.
Booster
Planned an app for uploading files in advance and collecting the finished prints. As planner and PM, I designed the user flow. I coordinated scope, priorities, and schedules for a 13-person team over three weeks to complete the project. It won the Grand Prize at the 16th APPJAM.
# skills
Tech stack
Backend
Web & Tools
Data
Data Pipelines
Infra
Observability
# education
Chung-Ang University · Graduated with honors · GPA 4.14/4.5 (95.89/100)
Chung-Ang University
B.S. in Industrial Security & B.S. in Software (double major)
Graduated with honors · GPA 4.14/4.5 (95.89/100)
- CS: Data Structures, Algorithms, Operating Systems, Computer Architecture, Software Engineering, Programming Languages, Network Application Design, Machine Learning, Probability & Statistics, Software Project
- Security: Information Security Theory, Security Data Analysis, Cyber Incident Response, Industrial Security Investigation & Forensics
Certifications
- Engineer Information Processing
- Jun 2021
- SQLD
- Dec 2021
- Big-Data Analysis Engineer
- Dec 2024
Languages
- Korean
- Native
- English
- Advanced
- Chinese
- Intermediate